To know where the birds are going, researchers turn to citizen science and machine learning

Computer scientists at the University of Massachusetts Amherst, in collaboration with biologists at the Cornell Institute for Ornithology, recently published in the journal Ecology and Evolutionary Methods A new prediction model that can accurately predict where migratory birds will go next. This is one of the most difficult tasks in biology. The model, called BirdFlow, is still a work in progress, but will be available to scientists later this year and eventually to the public.

“Mankind has tried to unravel bird migration said Dan Sheldon, professor of information and computer science at UMass Amherst, lead author of the paper and an avid amateur birder. UMass Amherst said, “It’s incredibly difficult to know exactly. real time information Not to mention which birds are where and where exactly they are going. ”

Many efforts, both past and present, have been made to tag and track individual birds, yielding invaluable insights. But physically tagging enough birds to form a good enough picture to predict their movements is difficult (not to mention the cost of such an undertaking). “It’s very difficult for tracking methods to understand how whole species move across continents,” says Sheldon.

In recent years, there has been a surge in the number of citizen scientists monitoring and reporting migratory bird sightings. Through eBird, a project managed by the Cornell Institute for Ornithology and international partners, birders around the world contribute to more than 200 million bird sightings annually.

It is one of the largest biodiversity-related scientific projects in existence, with hundreds of thousands of users, and promotes state-of-the-art species-occurrence modeling through the Lab’s eBird Status & Trends project. “The eBird data is amazing because it shows where a particular species of bird is every week across its entire range,” says Sheldon. – Level patterns.

BirdFlow takes eBird’s Status & Trends database and estimates of relative bird populations and runs that information through a probabilistic machine learning model. The model is tuned using real-time GPS and satellite tracking data so it can “learn” to predict where individual birds will move next.

The researchers tested BirdFlow on 11 North American bird species, including woodcocks, woodthrushes, and Swainson’s hawks, and found that not only did BirdFlow outperform other models for tracking bird migration, it also performed real-time We found that we could accurately predict the flow of movement without GPS. Satellite tracking data makes BirdFlow a valuable tool for tracking species that may literally fly under the radar.

“Today’s birds are undergoing rapid environmental change, and many species are declining,” says Benjamin Van Doren, a postdoctoral fellow at Cornell University’s Institute of Ornithology and co-author of the study. says. “By using BirdFlow, we can integrate different data sources to get a more complete picture of bird movement,” he adds Van Doren.