From declines to recovery: 52 years of changes in autumn migratory songbird abundance at an island stopover site in southern New England
Data files
Jun 26, 2025 version files 228.63 KB
Abstract
Migratory birds are declining across North America, and many studies have used long-term datasets to estimate trends in abundance for migratory birds. To most efficiently prioritize at-risk populations, it is critical to know the current trend status of migratory species, which involves accurately assessing the most appropriate timescale for monitoring changes in populations. In this study, we analyzed autumn hatch-year bird banding data from the Block Island Banding Station on Block Island, Rhode Island over a 52-year period (1970-2021) to examine long-term abundance trends for 22 species of migratory songbirds and compared our results to trends from Manomet Conservation Sciences and the USGS Breeding Bird Survey. We ran four models (no change, linear, quadratic, and breakpoint model) for each species and used model selection with AICc to determine the best model. Eighteen of the 22 species were best represented by breakpoint models, indicating that they experienced a sudden change in slope during the study period. For our 18 breakpoint species, all experienced the majority of their decline in the first two decades of the study (between 1976 and 1986), and 17 of 18 species were stable or recovering after the breakpoint. Of the four species without a significant breakpoint, one species had no change over the study period, two were best represented by linear models, and one was best represented by a quadratic model. Trend classifications varied across time, space, methods, and the geographic regions of the sub-populations sampled. These results present evidence that many species are currently stable or in recovery, despite initial declines, and that modern conservation efforts need to account for abrupt changes in trend direction within long-term time series analyses to most accurately assess current abundance trends for the conservation of vulnerable species.
Dataset DOI: 10.5061/dryad.h70rxwdwn
Description of the data and file structure
The provided R Scripts and data are used to analyze 52 years of autumn bird banding data (hatch-year birds only) from the Block Island Banding Station on Block Island, RI, USA. Specifically, this code runs four models (no change, linear, quadratic, and breakpoint) for 22 species of migratory songbirds, selects the best model, runs a leave-one-out analysis, and classifies each trend based on trend direction.
Files and variables
File: AOU_species_codes.csv
Description: This file contains taxonomic classifications for each species, downloaded from the American Ornithological Society.
Variables
- SPECIES_ID: Taxonomic order
- AOU: Assigned four-letter AOU code
- SPECIES_NAME: Common name
- spp: Scientific name
- order: Order number
File: Ornithological_Applications_Capture_Rate_Data_1970_2021.csv
Description: This file contains the 98% migratory window and associated captures, net-hours, and capture rates for 22 species of migratory songbirds (1970-2021).
Variables
-
BANDING_YEAR: The banding year for each entry
-
SPECIES_NAME: The common name for each species
-
AOU: The four-letter AOU code for each species
-
Start_Date: Start date of the 98% migratory window for a given species in a given year
-
End_Date: End date of the 98% migratory window for a given species in a given year
-
Start_Date_Julian: Start date of the 98% migratory window for a given species in a given year, expressed as the day of the year
-
End_Date_Julian: End date of the 98% migratory window for a given species in a given year, expressed as the day of the year
-
Length_Mig_Window: The total length of the 98% migratory window for a given species in a given year
-
Sum_Captures: The total number of captures for that species in that year, within the migratory window
-
Total_Net_Hours: The total number of net-hours accumulated during the migratory window (includes days where the station was open, even if a species was not captured that day)
-
Capture_Rate: Captures per unit effort. In this case, captures per 100 net-hours
-
Log_Capture_Rate: Log-transformed capture rate
-
Scale_Banding_Year: The scaled version of banding year (mean of 0, SD of 1)
Note: 98% migratory window is defined as the period of time where 98% of a given species passed through Block Island in a given year, removing 1% of captures on either side.
Code/software
This script was run on R version 4.4.0.
This analysis uses the following packages:
- Tidyverse: used for data manipulation
- broom: used for storing/tidying model output
- MuMIn: used for calculating AICc
- segmented: used for running breakpoint models
- conflicted: used to deal with package conflicts (optional)