No one can predict exactly where birds go, but this mathematical model gets close
When in doubt, go with the BirdFlow.
Migratory birds will soon be on the move. Starting in March, spring migration will be underway across North America as songbirds, shorebirds, waterfowls, and birds of prey return back to town. Although scientists know that these critters will soon be taking flight, they’ve long been trying to pin down what routes the birds will take to get back to the states from the tropics they’ve been overwintering in.
A study out this week in the journal Methods in Ecology and Evolution from a team at UMass Amherst and Cornell University describes a new strategy to predict the flight paths of these migratory birds using computer modeling and sighting data from citizen science platform eBird. And according to the researchers, the forecasting abilities of BirdFlow, as its called, are pretty accurate.
“It’s incredibly difficult to get precise, real-time information on which birds are where, let alone where, exactly, they are going,” Miguel Fuentes, the paper’s lead author and graduate student in computer science at UMass Amherst, said in a press release.
[Related: These new interactive maps reveal the incredible global journeys of migrating birds]
For example, scientists know that birds like American woodcocks will migrate each year from Texas and the Carolinas to the southern reaches of Canada. But they could take a number of routes to and from the two destinations.
Tracking tags can help provide some partial clues, but it’s hard to tag every bird of interest. And weather radar is good for visualizing bird movements in real-time, but can’t really give much information about which species are in the flock, or how individual birds are behaving. Moon-watching robots, on the other hand, are good for observing individual behaviors, but rely too heavily on cosmic timing. Besides, from year to year, birds may alter their routes. And like all wild animals, they’re inherently unpredictable.
To get a more accurate, live read on migratory birds, BirdFlow, the probability-estimating machine-learning model the team developed, uses information about weekly bird sightings and population distribution data from eBird for training. It was fine-turned with up-to-date GPS and satellite tracking data in order to predict where certain birds are headed next in their journey.
“BirdFlow models can be trained on any species, even those not tracked by eBird, as long as relative abundance models are available,” the researchers wrote in the paper. The model was tested on 11 species of North American birds like the American woodcock, wood thrush and Swainson’s hawk, and it outperformed other migration prediction models. Plus, it can correctly predict a bird’s flight path even without real-time GPS or tracking data.
According to the press release, the team will utilize a $827,000 grant from the National Science Foundation to further improve BirdFlow and prepare a software package for ecologists, which they expect to release later this year. The researchers are also working on a more visual interface based on these models to engage the general public with.
“In addition to the ecological questions investigated in our case study, samples from BirdFlow models can be used to study other phenomena such as stopover behavior and responses to global change,” the authors wrote. “Finally, BirdFlow can raise public awareness about biodiversity and ecosystem health by providing a tool for outreach to engage scientists, bird-watchers, policymakers and the general public.”