Dryad
Published January 9, 2024 | Version v1
Dataset Open

Heavy migration traffic and bad weather are a dangerous combination: Bird collisions in New York City

Authors/Creators

  • 1. Columbia University
  • 2. University of Canterbury
  • 3. New York City Audubon*
  • 4. Great Hollow Nature Preserve & Ecological Research Center*
  • 5. Cornell Lab of Ornithology

Description

Bird-building collisions account for 365-988 million bird fatalities every year in the United States alone. Understanding conditions that heighten collision risk is critical to developing effective strategies for reducing this source of anthropogenic bird mortality. Meteorological factors and regional migration traffic may influence collision rates but also may be difficult to disentangle from other effects. We used 5 years of bird collision counts in New York City to examine the influence of nocturnal weather conditions and bird migration traffic rates on collisions with buildings during spring and fall. We found that seasonally unfavorable winds and conditions that impede visibility are important factors that influence rates of bird-building collisions during both seasons. Specifically, northerly and westerly winds and low visibility in the spring, and southerly and westerly winds and low cloud ceiling height in the fall are associated with higher collision risks. Generally, these weather variables associated most strongly with increased collisions when nocturnal bird migration traffic was high, with the exception of low visibility in spring, which was predicted to triple collision rates compared to high visibility, independent of bird migration traffic. Although legislation to turn off unnecessary nocturnal lighting for the entirety of the migration seasons may be an ultimate goal, a proximate goal invaluable for reducing collisions will be predicting which nights will be of highest risk and using this information to determine when mitigation efforts could be most effective.

Notes

Funding provided by: Columbia University
Crossref Funder Registry ID: https://ror.org/00hj8s172
Award Number:

Funding provided by: National Science Foundation
Crossref Funder Registry ID: https://ror.org/021nxhr62
Award Number:

Funding provided by: Leon Levy Foundation
Crossref Funder Registry ID: https://ror.org/033hnyq61
Award Number:

Funding provided by: Lyda Hill Philanthropies
Crossref Funder Registry ID: https://ror.org/032sf2845
Award Number:

Funding provided by: Amazon Web Services
Crossref Funder Registry ID: http://dx.doi.org/10.13039/100008536
Award Number:

Funding provided by: Amon G. Carter Foundation
Crossref Funder Registry ID: https://ror.org/00xege567
Award Number:

Methods

This dataset includes five years of spring and fall data compiled from three sources: (1) bird collision data from NYC Audubon's Project Safe Flight, (2) bird migration traffic data processed from the KOKX radar station, and (3) historical weather data from LaGuardia Airport weather station. 

The bird collision data from NYC Audubon's Project Safe Flight was collected at a total of 27 buildings in NYC across 11 spring/fall monitoring seasons. Volunteer collision monitors walked a route consisting of 1 to 6 buildings once per day and, for each building, recorded their monitoring start and end times and the number of birds they found. 

Weather data are from the LaGuardia Airport weather station and provide local, proximate measurements of observed weather conditions at ground-level. Values included in this dataset are mean nocturnal (dusk-dawn) values of the following measurements: zonal (east-west) and meridional (north-south) wind components, cloud ceiling height, visibility distance, and air temperature. Due to a highly skewed distribution, we converted visibility distance into a categorical variable and considered visibility "low" when visibility distance was <10 km, "medium" when visibility distance was between 10 to 16 km, and "high" when visibility distance was >16 km. 

Bird migration traffic data was obtained and processed from the KOKX radar station. The migration traffic values included in this dataset are calculated nightly averages of migration traffic rate (# individuals/km/hr). We defined medium migration traffic as average migration traffic (x̄ = 9.46 x 102 individuals/km/hr in spring and 1.69 x 103 individuals/km/hr in fall), low migration traffic as x̄ - 1 SD (4 individuals/km/hr in spring and 9 individuals/km/hr in fall), and high migration traffic as x̄ + 1 SD (1.86 x 103 individuals/km/hr in spring and 3.42 x 103 individuals/km/hr in fall). 

For our study, we standardized cloud ceiling height, zonal wind component, meridional wind component, temperature, day of year, and average migration traffic variables to have a mean of 0 and a variance of 1 to aid model convergence. We did this using the scale() base R function, which uses the following equation: ((x – x̄)/SD), where x̄ is the mean and SD is the standard deviation. 

Files

collisions_migration_weather_spring_fall_dataset.csv

Files (976.9 kB)

Name Size Download all
md5:2c5f47a12bcb8d310c03727a198b7ede
973.1 kB Preview Download
md5:2bb9014567d5379a870d82d48efff8d6
3.8 kB Preview Download