Dataset DOI: 10.5061/dryad.pg4f4qs3m

Description of the data and file structure

We used long-term bird banding data collected during spring and fall migration at five stations across eastern North America: Appledore Island Migration Station ("AIMS"), Braddock Bay Bird Observatory ("BBBO"), Long Point Bird Observatory's Old Cut Research Station ("CUT"), Manomet Bird Observatory ("MBO"), and Powdermill Avian Research Center ("PARC"). These stations set out mist nets during migration seasons, and capture, collect data on, and release all the birds that fly into these nets. Each station uses standardized protocols to collect data in the same way over decades. They keep track of daily net effort, which is calculated as the number of nets open on a given day, multiplied by the number of hours each net was open; we used seasonal net effort in our analyses, which is the sum of daily net effort across a migratory season. We use data from spring and fall for all sites except AIMS, for which we only have data from spring migration.

We use four-letter species name abbreviations throughout our datasets, which refer to bird species common names as specified here:

Files and variables

File: species_abundances.zip

Description: We calculated species abundances for each species during each year at each site (AIMS, BBBO, CUT, MBO, and PARC) and season (spring and fall). Files in this compressed folder follow the naming convention "[SITE]_[season]_abundances.csv".

Variables:

• n: the number of individuals of each species captured in each year of data collection at the site during spring or fall migration

• Species: the four-letter abbreviation of the species

• Year: year of data collection

File: phenology_by_species.zip

Description: We quantified migration phenology for each species during each year at each site (AIMS, BBBO, CUT, MBO, and PARC) and season (spring and fall). We calculated the early, peak, and late passage dates for each species (the date by which 10 %, 50 % or 90 % of a species' total captures were made, respectively). We did not calculate passage dates for a species in a year in which fewer than 5 individuals of that species were captured. Some individuals may be captured multiple times within a season: to determine migration phenology, we removed from our dataset any same-season recaptures. Files in this compressed folder follow the naming convention "[SITE]_[season]_phenology.csv".

Variables:

• Year: year of data collection
• Species: four-letter abbreviation of species
• net.hours: the seasonal net effort at the site
• percentile: Values are 10, 50, or 90, referring to the day of the year by which 10 %, 50 % or 90 % of all individuals of a given species were captured.
• dates: the day of the year by which 10 %, 50 %, or 90 % of all individuals of the species were captured.

File: abundance_slope_matrices.zip

Description: These data files are matrices (row names are in the first column) for every site (AIMS, BBBO, CUT, MBO, and PARC) and season (spring and fall). Row and column names are four-digit bird species name abbreviations. Each cell contains the rate of change in relative abundance for each pair of species (with NAs along the diagonal). These values were calculated as follows: first, we calculated relative abundance for each pair of species as the sum of the number of captures of both species divided by the total number of captures of all species at the site during a given season and year. Second, we constructed a linear model for each pair of species, in which the response variable was their relative abundance and the predictor variables were year and net effort. We extracted the slope coefficients from each of these models for each species pair, which are the values contained in these matrices. The file names in this compressed folder follow the naming convention "[SITE]_[season]_abundance_slope.csv".

File: cooccurrence_slope_matrices.zip

Description: These data files are matrices (row names are in the first column) for every site (AIMS, BBBO, CUT, MBO, and PARC) and season (spring and fall). Row and column names are four-digit bird species name abbreviations. Each cell contains the rate of change in co-occurrence for each pair of species (with NAs along the diagonal). These values were calculated as follows: first, we calculated co-occurrence as the proportion of individuals of both species that were captured on a day when both species were captured. Second, we constructed a linear model for each pair of species, in which the response variable was their co-occurrence and the predictor variables were year and net effort. We extracted the slope coefficients from each of these models for each species pair, which are the values contained in these matrices. The file names in this compressed folder follow the naming convention "[SITE]_[season]_assn_slope.csv".

File: phenology_slope_matrices.zip

Description: These data files are matrices (row names are in the first column) for every site (AIMS, BBBO, CUT, MBO, and PARC) and season (spring and fall). Row and column names are four-digit bird species name abbreviations. Each cell contains the rate of change in phenological dissimilarity for each pair of species (with NAs along the diagonal). These values were calculated as follows: first, we calculated phenological dissimilarity as the difference (in number of days) between two species' peak migration passage dates. Second, we constructed a linear model for each pair of species, in which the response variable was their phenological dissimilarity and the predictor variables were year and net effort. We extracted the slope coefficients from each of these models for each species pair, which are the values contained in these matrices. The file names in this compressed folder follow the naming convention "[SITE]_[season]_phenology_slope.csv".

File: Mantel_results.zip

Description: We tested whether stopover communities have changed significantly during the time frame of the study. For this analysis, we used repeated Mantel tests to conduct pairwise comparisons of the average co-occurrence network of the first five years of the dataset with each subsequent year. We extracted the correlation coefficient of each Mantel test, which represents the similarity of each pair of networks. Files in this compressed folder follow the naming convention "[SITE]_[season]_MantelR.csv".

Variables:

• Year: the year of data collection
• mantel.r: the Mantel's R coefficient from a comparison of that year's co-occurrence matrix with a matrix of the average of the co-occurrence matrices from the first five years of data collection.
• net.hours: the net effort that year, during the given season
• hrs.diff: the difference in net effort between that year of data collection and the average net effort of the first five years of data collection

File: node_strength.zip

Description: We assessed how a species' abundance and relative phenological shifts relate to it's co-occurrence with all other species at the site. We calculated "node strength," which is the sum of a species' co-occurrences with all other species. The files in this compressed folder follow the naming convention "[SITE]_[season]_node_strength.csv".

Variables:

• Year: the year of data collection
• Species: the four-letter abbreviation of the species name
• node.strength: the sum of a species' co-occurrences with all other species that year
• n.individuals: the number of individuals of a species captured that year
• net.hours: the net effort of that year's migration season
• median.passage: the date by which 50 % of the individuals of a species were captured
• baseline.passage.date: the average of a specie's peak passage dates in the first five years of the dataset. ("Peak" passage date is the date by which 50 % of the individuals of a species were captured).
• species.shift: the difference (in days) between the species median.passage and baseline.passage.date
• overall.median.shift: the median species.shift across all species in a given year
• shift.difference: the difference between a species' phenological shift (species.shift) and the community's overall phenological shift (overall.median.shift).

File: proportion_foliage_gleaners.csv

Description: We tested whether the relative proportions of foraging guilds (birds using different foraging tactics) have changed over time. We only used the longest-term dataset for this question (from PARC). We selected three weeks of data from each season and year: a week of early migration, centered on the date by which 10 % of the season's birds were captured, a week of peak migration, centered on the date by which 50 % of the season's birds were captured, and a week of late migration, centered on the date by which 90 % of the season's birds were captured. Foraging categories (aerial forager, flycatching, foliage gleaner, and ground forager) are from Cornell Lab of Ornithology's "All About Birds."

Variables

• Year: year of data collection
• Aerial_Forager: the number of captured individuals that were aerial foragers
• Flycatching: the number of captured individuals that were flycatchers
• Foliage_Gleaner: the number of captured individuals that were foliage gleaners
• Ground_Forager: the number of captured individuals that were ground foragers
• total: the total number of individuals captured
• Proportion_Foliage: the proportion of total individuals that were foliage gleaners
• Season: the season of data collection (Spring or Fall)
• Week: the week of data collection (early, median, or late)

Co-occurrence arrays:

At each site, during both seasons, we calculated co-occurrence for each species pair. Co-occurrence was calculated as the proportion of individuals of both species that were captured on a day when both species were captured. For each year of data collection, we constructed a co-occurrence network: these are matrices in which every row and column refers to a species, and every cell contains that species pair's co-occurrence that year. These matrices are saved in a compressed folder, with one matrix for every year of data collection, following the naming convention: "association_matrix_[SITE][season][year].csv".

The first column in each dataset are the row names for the matrix, and the row and column names of each matrix are the four-letter species abbreviations. The diagonals are "NA." These matrices are stored in the following compressed files, following the naming convention "[SITE]_[season]_cooccurrence_array.zip"

File: AIMS_spring_cooccurrence_array.zip

File: BBBO_spring_cooccurrence_array.zip

File: BBBO_fall_cooccurrence_array.zip

File: CUT_spring_cooccurrence_array.zip

File: CUT_fall_cooccurrence_array.zip

File: MBO_spring_cooccurrence_array.zip

File: MBO_fall_cooccurrence_array.zip

File: PARC_spring_cooccurrence_array.zip

File: PARC_fall_cooccurrence_array.zip

Code/software

All analyses were conducted in R (version 4.5.0) and RStudio (version 2024.12.1+563).