From radar blobs to bird species, AI helps transform migration tracking

How can we identify specific bird species as they undertake their annual migrations?

In the U.S., scientists have long relied on weather radar data to track flocks of birds migrating across the country. But that came with a caveat: the birds appeared as blobs on the radar, much like dark clouds. They couldn’t tell what species of birds they were looking at.

A new tool is helping scientists and conservationists bridge this gap.

BirdFlow combines citizen science observations with weather data and artificial intelligence technology to help scientists detect specific species and predict their migration routes. Two studies published in the journals Global Ecology and Biogeography and Movement Ecology describe how BirdFlow can distinguish between species that are migrating across North America. The tool was developed as a collaboration between Cornell University, the University of Massachusetts, and the University of Illinois Urbana-Champaign.

“Let’s say we want to track Wilson’s warbler that winters in Alaska,” Yuting Deng, a co-author on both studies and postdoctoral researcher at the Cornell Lab of Ornithology, told Mongabay in a video interview. “Where exactly do they go? Do the eastern and western populations have similar migration timing? What routes are they taking? That’s where BirdFlow comes in.”

Knowing which bird species are migrating where is critical for several reasons.

For one, without being able to distinguish individual species, scientists can’t tell when and where a threatened species might be moving. The data are also useful for aviation managers, who would want to track larger birds that could potentially pose danger by colliding into aircraft. Additionally, there have also been campaigns across the U.S. to get members of the public to turn lights off at night when birds are migrating, in order to prevent them colliding into buildings. For this too, Deng said, it’s crucial to know what species are moving because “some bird groups are more prone to collide with buildings.”

To build BirdFlow, scientists tapped into a massive data set of citizen science observations from eBird, an app developed by Cornell. Over the years, a team at eBird has used the 2 billion observations recorded via the app to produce weekly maps of bird distribution and abundance. The team that developed BirdFlow used this data to develop what they call the BirdFlow Migration Traffic Rate, a metric that can estimate migration intensity at a species level on a weekly basis. The team validated the models by checking them against 30 years’ worth of data from 152 weather surveillance stations.

Additionally, they combined GPS, tagging and radio telemetry data to produce BirdFlow models for 153 migratory bird species in North America. The models help predict where the birds are likely to migrate from one week to the next, allowing scientists to map out migration routes.

“For BirdFlow, we just ask from week one to week two about the location of the birds,” Deng said. “For example, what is the probability of birds from one location flying into other locations adjacent to it.”

In addition to tracking threatened species and preventing building collisions, Deng said the data could also be used to monitor avian infectious diseases. Federal and state agencies have already expressed interest in BirdFlow data, she added.

“They want to see if we can create a forecast for specific wild bird species and how their migration range is overlapping with poultry farms,” she said. “If there is an outbreak in these farms, they can track where the birds could take these diseases to.”

The team also plans to create specific alerts for bird groups. Such alerts, Deng said, could possibly get the general public more excited and engaged if they know what species they’re helping protect on a given night.

More importantly, it could also be an active part of bird conservation and protection initiatives. “Different agencies like conservation groups and aviation managers can take the information that’s more targeted to the groups they want to conserve,” Deng said.

Banner image: A flock of snow geese. Image by Veronika_Andrews via Pixabay.

Abhishyant Kidangoor is a staff writer at Mongabay. Find him on 𝕏 @AbhishyantPK.

From caws to code: AI helps decrypt animal communication

Chewing sounds can help decode an animal’s diet using AI, new study finds

Citations:

Chen, Y., Slager, D. L., Plunkett, E., Fuentes, M., Deng, Y., Mackenzie, S. A., … Dokter, A. M. (2026). Population-level migration modeling of North America’s birds through data integration with BirdFlow. Movement Ecology, 14(1). doi:10.1186/s40462-026-00651-z

Plunkett, E., Deng, Y., Slager, D. L., Fuentes, M., Chen, Y., Van Doren, B. M., … Sheldon, D. (2026). Novel estimates of bird migration traffic at the continental scale using participatory science data. Global Ecology and Biogeography, 35(4). doi:10.1111/geb.70236

Cite this article

Abhishyant Kidangoor (2026). From radar blobs to bird species, AI helps transform migration tracking. Mongabay Conservation news. DOI: https://doi.org/10.66709/news-325488