Cloudy with a chance of warblers
Strides in computer power and artificial intelligence have enabled scientists to gauge the almost unimaginable scale of nighttime migrations—more than 700 million birds on some nights in the US.
By Warren Cornwall
On September 8, Neil Gilbert saw a forecast that made him set his alarm for 4:30 the next morning. On his computer, a map of the United States showed a bright yellow swathe of light over Madison, Wisconsin, where he lives. The prediction: More than half a billion birds would be migrating across the US that night, and the skies near his home were expected to fill with thousands of birds.
For decades, birders such as the 29-year-old University of Wisconsin ecology graduate student have looked to weather forecasts, calendars, and reports from other enthusiasts to track the spectacle of millions of birds flitting across entire continents. Now, he can tap into high-tech tools that are opening the door to a new understanding of bird behavior.
In just the past few years, a combination of weather radar, artificial intelligence, and cloud computing has enabled researchers to create a remarkably detailed, hour-by-hour picture of rivers of birds as they sweep across US skies each spring and fall.
The technology has given scientists new insights into the sheer numbers of birds migrating through the US—more than 700 million some nights. It is unlocking secrets about changes in migratory patterns wrought by habitat destruction, light pollution, and climate change. Forecasts predicting waves of birds days in advance herald a time when wind turbines might be automatically slowed or skyscrapers dimmed in advance of a migration surge. Scientists dream of a day when data streams become rich enough that they can simultaneously peer inside the brains of individual migrating birds and monitor the world around them, bringing people closer to unraveling the mysteries of life on the wing.
Migration forecasts days in advance create options for saving birds. Before each surge, wind turbines could be automatically slowed and skyscraper lights dimmed.
Almost from the moment radar was invented, birds showed up on it. In the beginning, they were little more than a headache. Radar (an acronym for “radio detection and ranging”) tracks objects by sending out radio waves and tracking objects that reflect those waves back. After its development by the British during World War II, the military quickly learned it needed to distinguish between passing geese and incoming German airplanes or warships. Two British biologists who served in the military during the war, David Lack and George Varley, described the confusion in a 1945 article in the journal Nature. There was only a hint at insights to come. “The radar results have produced little of ornithological interest,” they wrote. “Except to show how frequently birds fly at night over the sea.”
More than a decade later and across the Atlantic in New Orleans, Louisiana, Sid Gauthreaux, a teenage birding enthusiast, learned that a new weather radar system was being installed atop a downtown federal building. “I instantly, kind of brazenly, went up to the weather bureau and asked them if they could pick up birds on this system. They looked at me a little crazy,” recalled Gauthreaux, now a retired Clemson University behavioral ecologist recognized as a pioneer in using radar to study bird ecology.
In 1961, Gauthreaux, then a college student, had won the confidence of the local weather office and had confirmed, by comparing radar readings with observations of birds flying in front of the moon on clear nights, that the system could track birds. It was the beginning of his life’s work.
Situated on the Gulf Coast, Gauthreaux watched as masses of birds headed to Central and South America in the fall and then reversed course in the spring along what’s known as the Mississippi Flyway. The radar showed flocks of northbound birds exploding from coastal woodlands into the sky as night fell.
In the coming decades, advances in radar enabled Gauthreaux and others to gauge the almost unimaginable scale of nighttime migrations in the US, as well as the decline in migratory bird numbers across the Gulf. He made his first foray into forecasting future bird migrations, working with US Air Force bases to predict days when weather and bird-migration patterns raised the risk that low-flying fighter jets might collide with birds.
In 1999, the US Environmental Protection Agency commissioned Gauthreaux to lead the creation of BirdCast, a migration-forecasting tool meant to reduce pesticide poisoning of migrating birds by giving farmers advance warning. But computers of the era choked on masses of information from more than 140 radar stations around the country. Just getting the data required cumbersome paperwork.
“The vision was there,” said Andrew Farnsworth, then a graduate student in Gauthreaux’s Clemson lab. “But the reality was it took months to process half of the radar stations’ data to show a week’s worth of migration.”
That picture changed, starting in the mid-2010s. Cloud computing enabled the government to load masses of radar data onto easy-to-access networks. The growth in computer power, combined with the emergence of a form of artificial intelligence called machine learning, enabled scientists to find meaningful patterns in mountains of data and then use those patterns to predict future events.
The vision started to become reality. Today, the BirdCast website features maps of the US, with colors denoting places where larger numbers of birds are migrating—black for no major migration, purple for low levels, all the way to bright yellow for the hotspots. Orange arrows show which way the birds are flying. A box shows the number of birds predicted to be migrating on the current night, like a stock ticker tracking the Dow Jones.
On a recent night, the number stood at 340 million. Enter the name of a city into a search, and it provides a forecast for the migration levels at that location for the coming three nights. Another map allows a viewer to watch the night’s migration unfold in real time. Amid the fall migration season, the map shows bright colors washing across the continent from east to west—a sort of nightly sunrise as hundreds of millions of birds take to the air.
Behind the colorful graphics, computers churn through radar images looking for clouds bearing the trademark signs of birds, rather than rain. The system matches this with environmental factors including temperature, air pressure, humidity, and wind, and compares it to patterns found in more than two decades of radar readings.
“I kind of brazenly went up to the weather bureau and asked them if they could pick up birds on this system. They looked at me a little crazy.”
The site is addictive. Gilbert, the Wisconsin birder, checks the forecast multiple times a day, gauging when to grab his binoculars and head for the woods. As dawn broke on September 9, he stood with three friends at Bakken’s Pond, a patchwork of forest and marsh at the edge of the Wisconsin River. He could see trees alive with the movements of tiny songbirds—common jade-green Tennessee warblers mixed with zebra-striped, black-and-white warblers—and a real prize, the rarely spotted golden-winged warblers. The migrants were headed as far away as South America. The forecast had proven accurate. “I would guess this morning will probably be the most active morning I’ll see this fall migration,” he said later that day. “It was really quite hopping.”
The revelations extend far beyond recreational birdwatching. Radar-wielding researchers have dissected how wind patterns influence migrating birds in flight, learning that birds adjust altitudes to maximize the benefit from the wind, drifting off the most direct route to adjust to crosswinds. They’ve measured how migration unfolds in pulses, with more than half of all US migratory birds on the move during just 10 percent of the months-long spring and fall migration seasons.
Others have created detailed maps highlighting critical stopover spots for migrating birds. Radar revealed that as many as 2 billion birds take rest breaks each spring and autumn in coastal forests in the southeast, in places such as the Big Bend area in northern Florida.
Radar-powered research isn’t confined to North America. In northern Europe, scientists are working to build an automated tracking system across the continent. Chinese and Australian scientists are wielding radar to track crop-destroying insects such as locusts. But when it comes to birds, US scientists have the advantage of a single network of government weather radar covering much of a continent. “There are people all around the world working on this. None of them have quite gotten to the level we have, mainly because of the lack of access to data,” said Farnsworth.
The technology has also painted a portrait of how humans are tinkering with migration. Rising temperatures in the last two dozen years have pushed the peak of migration earlier, at the rate of around half a day per decade. Using radar and microphones that pick up bird calls, scientists illustrated the effects of artificial light pollution by revealing that powerful lights beamed into the sky from the 9/11 memorial in Manhattan attracted and disoriented migrating birds, causing them to slow and fly in circles. In Chicago, researchers using radar paired with tracking of dead birds around a single tall building concluded that halving the number of lit windows at night would cut bird deaths there by 60 percent.
Kyle Horton hopes these new insights could help protect birds as they navigate this increasingly hostile landscape. The Colorado State University ornithologist, who once worked in Farnsworth’s Cornell lab, has a hand in much of this research. He envisions a time when alerts about a coming wave of migratory birds could be used to adjust wind turbines or turn down the lights on skyscrapers. “If this area is going to see very high amounts of migration, then we could deploy some conservation action,” he said.
That’s already happening in some places, at a simple level. Spurred by a grant, the Cornell-based BirdCast program last year started working with conservation groups to encourage Texas cities, building owners, and residents to turn off lights during peak migration days. In Dallas, Mayor Eric Johnson in 2020 urged people to take part between September 19 and October 10, a time period highlighted by the radar-based forecast. He issued a similar declaration in the spring.
Having a narrow window of peak migration has helped persuade some building owners who are reluctant to turn off night lights for months at a time, said Ben Jones, executive director of the Texas Conservation Alliance, a Dallas-based environmental group leading the campaign there. In the spring, 23 building owners signed on, and he has seen city hall and the city library go dark. “The good thing is, the campaign allows them to participate in whatever capacity they can afford,” he said.
Farnsworth, meanwhile, said there is interest in the wind industry about ways to incorporate such migration warnings into turbine management to minimize the time turbines are shut down while maximizing the benefits for birds. “I know that some in the wind energy industry are thinking very seriously about this.”
In the next five years, he foresees scientists weaving together the growing mass of data to create a richly detailed tapestry of unfolding migrations. Radar depicting clouds of birds would be supplemented with daily reports of bird sightings, using apps such as eBird and acoustic recordings to help pinpoint which species are on the move.
In the distant future, Farnsworth wonders whether there will be a time when radar tracking of weather and mass migration will be paired with sophisticated sensors mounted on individual birds, to measure their movements, brain activity, and body chemistry. “I’ve been thinking a lot about how you could use that information,” said Farnsworth. “There’s brain processing going on, so far beyond or so different than our capacity.”
The results could bring scientists closer to a true bird’s-eye view of the world.
Warren Cornwall is a freelance journalist based in Bellingham, Washington. His work has appeared in The New York Times, Boston Globe Magazine, and the Seattle Times. He is a contributing correspondent for Science.
What to Read Next
Decades of declining herbivore populations in central African national parks have conservation scientists arguing that smaller reserves might be better for wildlife
New research suggests trees in the future won't get a big growth spurt from more CO2, and will die more in wildfires, droughts and insect outbreaks.