Data and code from: Bioacoustic monitoring reveals patterns of landscape use by migrating birds at a Great Lakes barrier crossing
Data files
Jan 04, 2026 version files 4.80 MB
-
Keweenaw_Reoriented_Migration.zip
4.80 MB
-
README.md
6.11 KB
Abstract
Understanding how highly mobile animals use landscapes at broad geographic scales remains a major challenge in ecology. Traditional monitoring approaches often lack the spatial and temporal resolution to monitor how migratory species use heterogeneous landscapes, contend with movement barriers, and interact with urban and developed landscapes. Here, we use a passive acoustic monitoring network to characterize landscape use of migrating songbirds in the Keweenaw Peninsula, a major barrier crossing point along the south shore of Lake Superior. Using nearly 3 million acoustic detections of migrants from 18 sites spanning 328 km2, we demonstrate that landscape use is shaped strongly by local geography and wind conditions during reoriented movements associated with barrier crossing. Generally, more songbirds used the peninsula during wind conditions favorable for migration. Following winds unfavorable for crossing in spring, birds concentrated in coastal and ridge landscapes oriented along an east-west axis. Geographic gradients and coastline orientation both played important additional roles in shaping migrants’ landscape use. Together, our results illustrate the complex role of large water barriers in shaping landscape use in highly mobile animals. More broadly, our findings demonstrate the value of acoustic monitoring as a novel technique for studying migratory animals’ landscape use. This approach offers a powerful, scalable tool that can be deployed across complex landscapes and inform conservation priorities of hard-to-monitor species.
Dataset DOI: [10.5061/dryad.5mkkwh7hq]
Description of the data and file structure
Data sources for: Bioacoustic monitoring reveals patterns of landscape use by migrating birds at a Great Lakes barrier crossing
Authors: Zach G. Gayk and Benjamin M. Van Doren
Author contact: University of Illinois Urbana-Champaign, Department of Natural Resources and Environmental Sciences, Urbana, IL 61801, USA
(1) zachgayk@gmail.com (2) vandoren@illinois.edu
Data were collected by using acoustic monitoring units to detect migratory songbirds' flight calls between 2022-2024 at 12 sites in the Keweenaw Peninsula. Michigan, USA. This peninsula concentrates migratory birds attempting to cross Lake Superior in the Great Lakes of North America. The Nighthawk machine learning algorithm was used to detect and identify songbird flight calls in the dataset. Subsequent manual review of randomly sampled Nighthawk detections was used to filter Nighthawk detections to only include high-confidence data.
Files and variables
Files: (1) merged_three_year_dataframe_2_December_2024.txt
(2) total_combined_seasons_PCs_spring_with_log_radar_3_December_2024.txt
(3) total_combined_seasons_PCs_fall_with_log_radar_11_December_2024.txt
Each file is found within the data folder of the Keweenaw_Reoriented_Migration.zip project folder.
Description:
Variables
- night_of: Night of migration
- time_bin: period of Night/Day from which wind data were averaged
- mean_u: mean east-west component of wind (m/s) where negative is east and positive is west
- max_u: max east-west component of wind (m/s) where negative is east and positive is west
- mean_v: mean north-south component of wind (m/s) where negative is north and positive is south
- max_v: max north-south component of wind (m/s) where negative is north and positive is south
- det_start_date: Calendar date of detection
- predicted_category: Nighthawk taxonomic category
- lat: Latitude
- lon: Longitude
- site_id: Site Id of recorder
- observation_count: total daily count of each predicted category
- season: Season
- year: Year
- night_of_numeric:
- mean_u_6am_to_11am_next_day: mean U in this time period for each day
- mean_u_9pm_to_midnight_prior: mean U in this time period for each day
- mean_u_midnight_to_5am_next_day: mean U in this time period for each day
- mean_u_noon_to_5pm_next_day: mean U in this time period for each day
- mean_v_6am_to_11am_next_day: mean V in this time period for each day
- mean_v_9pm_to_midnight_prior: mean V in this time period for each day
- mean_v_midnight_to_5am_next_day: mean V in this time period for each day
- mean_v_noon_to_5pm_next_day: mean V in this time period for each day
- diff_mean_u_9pm_to_midnight_prior_6am_to_11am_next_day: the difference in mean U between each time period in delta m/s where negative is increasing west wind and positive is increasing east wind
- diff_mean_v_9pm_to_midnight_prior_6am_to_11am_next_day: the difference in mean V between each time period in delta m/s where negative is increasing south wind and positive is increasing north wind
- diff_mean_u_9pm_to_midnight_prior_midnight_to_5am_next_day: the difference in mean U between each time period in delta m/s where negative is increasing west wind and positive is increasing east wind
- diff_mean_v_9pm_to_midnight_prior_midnight_to_5am_next_day: the difference in mean V between each time period in delta m/s where negative is increasing south wind and positive is increasing north wind
- radar_KMQT: nightly migration from the KMQT radar
- mt_KMQT: nightly migration from the KMQT radar
Variables only found in total_combined_seasons_PCs_spring_with_log_radar_3_December_2024.txt and total_combined_seasons_PCs_fall_with_log_radar_11_December_2024.txt
- PC1 (response variable summarizing overall daily intensity of avian migrant landscape use across the Keweenaw peninsula)
- PC2 (response variable summarizing daily northwest-to-southeast (spring) and west-to-east (fall) gradients in the intensity of avian migrant landscape use across the Keweenaw peninsula)
- PC3 (response variable summarizing daily north-to-south gradients in the intensity of avian migrant landscape use across the Keweenaw peninsula)
Code/software
Code was produced in R version 2023.12.1+402
All code consists of .R files.
main dataset consists filtered Nighthawk detections of flight calling migrants with mean U and mean V wind,
averaged across nightly time periods. This dataset is used to produced separate spring and fall datasets with daily PC Scores representing intensity of migration
across 18 sites in the Keweenaw Peninsula, Michigan.
Dataset name: merged_three_year_dataframe_2_December_2024.txt
additional datasets consist of
(1) total_combined_seasons_PCs_spring_with_log_radar_3_December_2024.txt
and (2) total_combined_seasons_PCs_fall_with_log_radar_11_December_2024.txt
these two datasets consist of mean U and mean V wind data across several time periods of the night, nightly radar traffic over Lake Superior, and daily PC scores summarizing bird migration
for each day. Spring and fall data were analyzed separately with unique PCs for each season. These two datasets were used to run models on the influence of wind, nightly migration traffic (radar), and Date of season on daily PC scores representing diurnal migration and landscape use.
Code:
total_analyses_final_9_December_2025.R contains code to conduct PCA on migratory bird landscape use and to produce models of migratory landscape use versus wind data
Final_paper_figures.R contains code to produce figures used in main manuscript based on partial residual plots from models of PC1, PC2, and PC3 each used separately as response variables of intensity of habitat use throughout the Keweenaw Peninsula
all code is within the scripts folder of the Keweenaw_Reoriented_Migration.zip project folder.
