ABSTRACT

Billions of birds migrate each year in journeys that are largely hidden from human observation, yet are
critical to the success of bird populations. To understand and monitor migratory species, data and
methods are needed that can capture the movements of bird populations across the globe. The eBird
citizen science project receives millions of bird observations throughout the year and uses these data to
produce detailed weekly abundance maps for hundreds of migratory species around the world. Despite
this rich information about bird distributions, scientists lack widespread, detailed data about the
migratory routes that link bird populations and their habitats throughout the year. In the BirdFlow
project, a team of computer scientists and ornithologists will use citizen science data to create models
and algorithms to infer population movements of migratory birds. The models will allow inferences
currently unavailable to ecologists at the scale of full populations and flyways, including simulated
migration routes and movement forecasts. The resulting data will help address urgent needs in ecology,
conservation, and industry, including understanding connectivity between populations and links
between migration and evolution, as well as applications to disease spread and aviation safety.
Visualizations and educational material will be created to inspire the public and raise awareness about
biodiversity and ecosystem health.

The BirdFlow project will develop models and algorithms to infer bird movements from citizen science
data. Data products from the eBird Status and Trends project will provide information about the weekly
distributions of bird populations, and optimization problems will be formulated to infer population
movements that are consistent with the weekly distributions and approximately minimize energetic
costs. Individual tracking data and other evidence will be used to validate and improve models.
Technically, the work will build on an emerging line of research that uses probabilistic graphical models
to learn about probability distributions over many variables from partial information, such as noisy
estimates of the distributions of individual variables. Software and data products will be created that
will allow scientists to use pre-fitted BirdFlow models to simulate synthetic migration routes and create
movement forecasts for species of interest. The project team will use BirdFlow to conduct ecological
research about patterns and drivers of migration in the Western Hemisphere. Project information can
be found at https://birdflow-science.github.io/.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

Please report errors in award information by writing to: awardsearch@nsf.gov.