eBird Takes Data to the Stars, or Satellites at Least

The leading science journal Nature has an article today about eBird working with satellites and supercomputers. It’s a nice explanation of a new development that the eBird team (a joint project of the Cornell Lab and Audubon) is really excited about: they’ve been awarded 100,000 hours of computing time on the National Science Foundation’s supercomputers.

What would we possibly need so much supercomputing time for, you might wonder. It’s not just that eBird, now in its ninth year, gets upwards of a million entries per month—our own servers do a pretty good job of storing those data and serving it back to the public (so that users like you can generate your own lists or view global range maps). More than that, as director Steve Kelling told Nature, his group is taking up the challenge to turn this bonanza of birding observations into something meaningful to science.

To do that, the eBird team is working with a group of collaborators (see below) to match bird observations with data about land cover and vegetation from satellites. Those relationships can be used to model where and when individual species move across the country, as in the above animation of Eastern Phoebes moving around the country over the course of a single year. (There’s a similar animation where you can see Indigo Buntings taking over North America in spring, on the Nature site).

These maps are deceptively simple. They represent the combined knowledge of thousands of birdwatchers—the kind of information that used to be stowed away in thousands of notebooks, or as checkmarks on checklists—now brought together, analyzed, and used to estimate a real, living range map for the species that changes with each week of the year. And Kelling’s plans include yet another step—using these relationships to explore likely scenarios about how climate change will affect bird distribution. And that’s where the supercomputers come in—those initial models took five days to run on our “normal” computers. Read the Nature story for more details.

This work accesses the full eBird database, consisting of almost 50 million bird observations from more than 500,000 locations across North America. The work is being done through collaborations with the Institute of Computational Sustainability, the Cornell University Department of Computer Science, DataONE at the University of New Mexico, and the Oak Ridge National Laboratory, Kelling said. Key people working on the project include the Cornell Lab’s Daniel Fink, Wesley Hochachka, and Kevin Webb, as well as Theodoros Damoulas, Carla Gomes, Bill Michener, Bob Cook, Suresh SanthanaVannan, and John Cobb at partner institutions. The work is being funded by the National Science Foundation, the Leon Levy Foundation, and the Wolf Creek Foundation.

We were also impressed that the Nature reporter managed to throw in a vignette from the Gulf Coast to kick off the story—it features Louisiana State University ornithologists Steve Cardiff and Donna Dittmann exploring unnamed islands, breaking a leg in the process, and persevering to enter their hard-won data into eBird (OK, now I feel bad for claiming I haven’t had time to enter my own eBird check-lists). By the way, that’s the same Steve Cardiff who helped me learn to survey Wilson’s Plovers in June; Dittmann was on the boat that same day, still wearing a walking cast from the injury the Nature reporter described.