- Since its launch in 1999, the BirdCast project has used weather radar data to track and forecast bird migrations across the U.S.
- In recent years, technology such as cloud computing and machine learning have helped make the work of researchers in the project easier and more automated.
- Studying bird migrations is essential not only to help protect them, but to also analyze and understand environmental health.
- The BirdCast project is now working on integrating radar data with human observations and bioacoustics to help identify the bird species traversing the skies.
Ask Andrew Farnsworth about the origins of the BirdCast project, and he would describe it as “ancient history.” The year 1999, when the project was conceived, is a far cry from antiquity. But given how technology has helped transform the project since then, Farnsworth’s description might not be completely off the mark.
BirdCast, in essence, is a project that tracks and forecasts bird migrations across the U.S. It uses weather radar data, in which birds also show up, to spot birds. Back when it was launched, scientists had the tedious task of manually poring through the chaotic radar data to differentiate between birds and meteorological phenomena like clouds.
“I used some statistical models and did some math that connects weather to migration intensity to make the forecast,” Farnsworth, senior research associate at the Center for Avian Population Studies at the Cornell Lab of Ornithology, told Mongabay in a video interview. “The state of the art then was being able to use what was in the human brain to make the forecast.”
Since then, BirdCast has evolved to combine ornithology and advanced computer sciences to understand and predict bird migration patterns. The project now uses technology such as cloud computing and machine learning, which have helped make the work of researchers easier and more automated. The team at the Cornell Lab of Ornithology is also currently working on combining human observations with technology like machine learning and bioacoustics to identify the species of migratory birds — information that has remained elusive so far in the radar data.
BirdCast’s transformation from the rudimentary project that it was in 1999 to its current sophisticated version, in fact, closely tracks the growth and evolution of artificial intelligence, specifically machine learning.
“When machine learning was starting to become a reality in the 2010s, we started thinking about how to reconfigure the project and take it into the 21st century,” Farnsworth said. “Now, we are reaping the fruits of all that because we have enormous amounts of data that we can interpret, analyze and visualize in an automated manner.”
As the project evolved, so has its applications.
BirdCast was initially launched to alert farmers about bird migrations to ensure they wouldn’t spray harmful pesticides or chemicals when birds were passing through. It was launched as a collaborative effort between several organizations, among them the Cornell Lab of Ornithology, the Clemson University Radar Ornithology Laboratory, and the U.S. Environmental Protection Agency.
Since then, researchers as well as governments have used the data obtained from the project to plan and implement conservation actions aimed at reducing bird mortality. For instance, New York City launched the “Lights Out” initiative, which used BirdCast data to understand the intensity of bird migrations at different times of the year. Subsequently, the initiative directed buildings owned and managed by the city of New York to turn off their lights at night during peak bird migration seasons in the fall and spring. The initiative aims to reduce bird mortality caused by collisions with buildings, which in turn is caused by artificial lights that attract and disorient the birds. According to a 2014 study published in the journal Ornithological Applications, at least 100 million birds die every year in the U.S. because they fly into buildings.
For scientists, understanding bird migrations also helps with understanding global warming and climate change. Birds are considered good indicators of environmental health, and are useful for assessing the health of ecosystems. Migration, Farnsworth said, is a particularly good metric because birds cue on environmental health, and any changes would impact their migration patterns.
Now, technology is aiding BirdCast’s work of analyzing weather radar data and deducing meaningful patterns from it. Twenty-five years ago, the U.S. weather radar data had to be downloaded manually; today, it’s automatically transmitted to a cloud computing platform. There, the machine-learning algorithms are activated to spot and identify birds or masses of birds. The filtered and visualized data then flow into BirdCast’s website, where they show up as live migration maps.
Machine learning is also helping scientists unlock decades’ worth of data to forecast bird migrations. Farnsworth and his team have trained the algorithm to analyze historical data from the past 25 years to do this work.
“We analyze all of that hundreds of millions of radar scan data to understand what the intensity of migration was like then,” he said. “The algorithm deduces patterns from weather variables like temperature, wind speed, direction and atmospheric pressure to predict the migration density.”
There’s likely more to come, too.
For now, scientists are unable to identify specific species of birds from the weather radar data. This means that while they can spot masses of birds flying by on the radar, it’s difficult to figure out what birds they are. Farnsworth and his team are now working to bring additional data sets to help solve this problem. This includes human observational data gathered by ornithologists, bird-watchers and citizen scientists over the years. The team also plans to incorporate data gathered using the tool of bioacoustics, which has become increasingly popular in recent years because of its ability to gather copious amounts of audio data over large landscapes.
Integrating these vast and diverse data sets, and finding out how they intersect with each other, is now at the core of the mission at BirdCast.
“How do you integrate all these different sources of information to put a face to the birds we spot in radar data?” Farnsworth said. “It is this integration that is going to be the biggest leap forward in the next five to 10 years.”
Banner image: A flock of snow geese. Image by Veronika_Andrews via Pixabay (Public domain).
Abhishyant Kidangoor is a staff writer at Mongabay. Find him on Twitter @AbhishyantPK.
Loss, S. R., Will, T., Loss, S. S., & Marra, P. P. (2014). Bird–building collisions in the United States: Estimates of annual mortality and species vulnerability. The Condor, 116(1), 8-23.