In a new study harnessing the power of tens of millions of citizen science observations, researchers found that fire repels some birds while others are drawn by the flames.
The birds’ responses also varied throughout their range. “Fire conditions can lead to high bird abundance in one region, but low bird abundance in another region for the exact same species,” said Andrew Stillman, an applied quantitative ecologist at the Cornell Lab of Ornithology and a coauthor on the study published in Frontiers in Ecology and the Environment.
Some species, like the red-cockaded woodpecker (Leuconotopicus borealis), always increased after a fire, albeit to different amounts throughout their range. But the researchers were surprised to see that for other species, like the American goshawk (Astur atricapillus), fires in one region led to an increase in abundance where in another area it led to a decrease.
Models all the way down
Alongside the eBird team at Cornell, Stillman creates analytical tools that help organizations like the U.S. Forest Service use eBird data to understand how birds and their habitats are changing through time. The agency was interested in broader scale information about the impacts of fire on wildlife, a task which would require massive amounts of data and the technical know-how to work with it.
The U.S. Forest Service partnered with the eBird Status and Trends team, a group of computer scientists, statisticians and bird scientists who have used eBird data to create maps for nearly 3,000 bird species around the world.

The team looked at six bird species of conservation concern. Along with the red-cockaded woodpecker and American goshawk, they tracked the Bachman’s sparrow (Peucaea aestivalis), greater sage-grouse (Centrocercus urophasianus), pinyon jay (Gymnorhinus cyanocephalus), and olive-sided flycatcher (Contopus cooperi).
They started with around 31 million eBird checklists from the contiguous U.S. during the breeding season from 2011 to 2021, with data quality filters established by the eBird team. “We only include the highest quality checklists that have both detection and non-detection information,” Stillman said. The authors then combined the eBird data with dozens of other variables describing land use, habitat type and fire history to run hundreds of machine learning models.
The models were so complex and numerous the team used servers hosted by the National Science Foundation’s Advanced Cyberinfrastructure Coordination Ecosystem: Services & Support (ACCESS) to complete the work.
Shedding light on forest management
The results showed birds didn’t respond uniformly to fire across their range. This is evidence against a “one size fits all” approach for managing fire to promote biodiversity.

“Managing forests and fire and biodiversity can be like driving at night,” Stillman said. Streetlights—analogous to local field studies—offer pinpoints of light in the darkness, but they usually can’t illuminate the overall landscape.
The study’s results can help inform fire management strategies by showing broad-scale patterns beyond just pinpoints of light. “If managers have information about biodiversity responses to fire at their fingertips, it will be easier for them to incorporate wildlife science into decision making,” Stillman said.
Participatory science data can also help cash-strapped agencies balance slim budgets. “It takes time—and a whole big team—but projects like this can represent huge costs savings for agencies,” Stillman said.
This first paper was a proof-of-concept. Now Stillman’s team and U.S. Forest Service biologists are looking at fire responses for more than 100 additional species.
“How cool is it that the key that finally unlocked this new information source is data collected by passionate wildlifers and birders from around the world,” he said.













