Four Amazing Impacts of This A.I.-Powered Bird Migration Tracker
A reimagined tool called BirdCast is helping birders, scientists and even farmers
This fall, when you step out of your house at night, millions of birds may be flying overhead. As the weather cools, hundreds of North American bird species make seasonal journeys from their breeding grounds to wintering locations farther south. An estimated four billion birds pass from Canada into the northern United States each fall, and almost five billion cross out of the southern U.S. into Mexico and beyond. But you might not notice this epic journey at all, because the dense flocks move under the cover of darkness.
For a long time, scientists couldn’t say for sure the birds were even taking a journey. In the 1800s, European naturalists could not explain how local birds disappeared during the winter. They first got an inkling that birds migrated long distances when a stork showed up in a German village in 1822 with a thin spear through its neck. The so-called arrow stork, or pfeilstorch in German, was one of 25 storks that have survived an attack while living in Central Africa and then turned up in Europe. Identifying the origins of the arrows eventually led to theories about long-distance migration.
Today, birds are “tagged” in less invasive ways that help scientists follow their movements. Researchers attach bands to bird legs or small radio devices to avian backs. But those methods still only capture small parts of populations. And for a long time, to measure migrations of large numbers of birds, scientists had to aim telescopes at the full moon and make careful tallies of silhouettes passing over the orb.
But thanks to gradual advances in artificial intelligence modeling and computing power, in 2018 a team of researchers at Cornell University boosted the capabilities of a tracking technology called BirdCast to monitor and predict mass movements of birds on the wing. The original project began in 1999, but the technology lacked automation and required hundreds of hours to analyze and predict migration. The platform now monitors birds as they glide over the U.S. using radar from the National Weather Service. A.I. models trained on years of historical data pull out live bird migration data in minutes and forecast their movements at the county level up to three days in advance. Five years since the innovative tool’s update, we checked in with researchers to hear how it is helping with science and conservation.
Forecasts alert cities to turn off their lights
Artificial light can draw migrating birds in and make them lose their sense of direction. As a result, the creatures may waste precious energy circling light sources and fall exhausted from the sky. Nighttime illumination also increases the rate of birds hitting buildings. Hundreds of millions of birds die every year when the animals fly into our structures, especially those that are made of glass or lit up.
But BirdCast may aid conservationists in bringing down the number of impacts. BirdCast maps imminent bird migrations in striking visuals. City and state Lights Out campaigns from Connecticut to Colorado share updates on bird movements and relay forecasts from BirdCast to mobilize local governments and business organizations to turn off lights. Julia Wang, project leader for BirdCast at the Cornell Lab of Ornithology, says she thinks BirdCast has motivated more people to rally around such actions.
“Migration season is long, and when you’re asking people to commit for months on end, it can be hard to get their attention,” says Wang. With the new and improved BirdCast, Wang says her campaigns can focus on critical windows of migration, when the highest number of birds are migrating and at risk of death related to light pollution. BirdCast can email tailored alerts to business owners or homeowners in 44 cities to mobilize residents to turn off lights and reduce bird fatalities.
Wang says BirdCast visualizes the scale and reality of migration, creating a feeling of urgency that can translate into action.
Dustin Partridge, director of conservation and science at the New York City Audubon, says his organization uses BirdCast every night during migration season to figure out how to prepare their team of over 100 volunteers that sweep the city looking for injured or dead birds. On nights with higher migration, the volunteers prepare with extra supplies and ready themselves for a longer shift. Project Safe Flight, an Audubon program that Partridge leads, treats birds that have been injured after crashing into buildings. The team also tallies strikes and fatalities.
Project Safe Flight’s members also monitor the 9/11 memorial called Tribute in Light, with volunteers counting how many birds stray into the glow. Officials turn off the two high-powered beams when too many birds become trapped in their light. Partridge says BirdCast predicts when waves of migrating birds will pass through and alerts officials to times when the lights may need to be turned off.
Predictions tip off birders about incoming flocks
Spectators can check BirdCast to see estimates of the number of birds approaching or currently flying through a county. Birders can inspect the speed, direction and altitude of the migrants. The data shows how current migrations stack up against previous nights or even previous seasons. And birders can use information from the night to inform birding the next day. For example, a 2022 Audubon magazine article advises that if a group of birds drops in altitude overnight, more migrating birds may be visible after the sun rises.
Andrew Farnsworth, a migration ecologist at the Cornell Lab of Ornithology who has worked on BirdCast for years, says he hopes the project will encourage new people to go out in nature to see migrating animals for themselves. “It’s about people going out into the field and observing what’s happening and maintaining the connection [to the birds], whatever we can do to enhance that connection.”
Birders, in turn, are helping improve BirdCast with their observations. Every time a birder logs a bird sighting in a publicly available database, BirdCast uses that information to create predictions about which species of birds will be migrating in a given place and at a given time.
Forecasts can warn farmers to guard against avian flu
Avian flu commonly spreads among wild bird populations, but a recent outbreak in 2022 grew into one of the largest flu epidemics in the U.S. poultry industry. According to the Centers for Disease Control and Prevention website, this recent strain of bird flu has led to the deaths—direct or indirectly through culling—of almost 59 million birds across the country. Shortages of egg-laying hens led the cost of eggs to more than double in 2022.
Yuko Sato, a poultry veterinarian and diagnostic pathologist at Iowa State University, says some poultry farmers have been looking at migration forecasts like BirdCast or observational logs like eBird and “hunkering down” with increased protections when wild birds are on the move. She adds that BirdCast and other bird tracking might eventually serve as part of a more formal early warning system if it were used to track birds during winter flu season as well as migration season: For example, if a bird flu outbreak erupts in Texas right before a northward migration of wild birds, farms in Iowa might get messages about when to be on high alert.
BirdCast might also be used to understand what drove previous flu outbreaks. In 2022, the U.S. Department of Agriculture compared BirdCast movement records with bird influenza rates. Their report suggested that periods of intense migration match up with increases in domestic poultry infection rates for the 2021 and 2022 season.
Researchers have turned data into dozens of insights
An academic project at its core, BirdCast is still being used to study birds. According to its website, the team that runs BirdCast at the Cornell Lab of Ornithology has published two dozen scientific articles related to the platform since its 2018 update. Some researchers have used the system to investigate the factors that put birds most at risk of hitting a building, or to potentially make aviation safer by building models of when planes hit birds.
Jeff Kelly, an avian ecologist at the University of Oklahoma who is not involved in BirdCast, says the project is remarkable, resting on the tireless work of relatively few scientists. The project lists only 15 current members on the website. Since its origins as a coarse forecasting tool in 1999, BirdCast has grown to include up-to-the-minute updates on bird movements in counties across the continental U.S., as well as city-level alerts that rely on A.I. forecasting tools. “I think it’s demonstrated value at a national level that goes beyond what we should expect to come out of one research lab,” Kelly says.