By Emma Marris
Midway through a birding expedition last May off the Louisiana coast, Donna Dittmann lost her footing and broke her leg. Unaware of this, she kept the weight off her swollen ankle while surveying birds the next day at an unnamed islet that was packed with nesting pelicans, egrets and terns. After she returned home, a visit to the emergency room revealed the extent of her injury. Her husband, Steven Cardiff, who, like Dittman, is a collections manager at the Louisiana State University Museum of Natural Science in Baton Rouge, dubbed the islet "Fractured Fibula Island" in her honor. He then he went online and added the species they had seen there to "eBird."
A database that records the vast numbers of sightings routinely made by dedicated birders around the globe, eBird has been growing steadily since its launch in 2002. More than 48 million observations have been entered so far--10 million of them in 2010 alone. The data represent millions of hours of eye-straining--and sometimes leg-breaking--observations.
According to Steve Kelling, director of information science at the Cornell Laboratory of Ornithology in New York, who runs the project together with the New York-based National Audubon Society, "the challenge now is to try to do something meaningful with all these data."
Fortunately, eBird has just been given some powerful help. Last week Kelling learned that the project has been awarded 100,000 hours on the National Science Foundation's TeraGrid supercomputer. By performing intensive data analysis using the supercomputer, Kelling and his colleagues hope to turn the scattered observations of each bird species into a global view of its movements.
The eBird team will start by combining the bird sightings with remote sensing information from sources such as the Moderate Resolution Imaging Spectroradiometers (MODIS) on board NASA's Terra and Aqua satellites. Among the data that can be gleaned from MODIS is precisely when different places on Earth are "greening up" in the spring--a seasonal phenomenon that can be strongly correlated with bird movement.
The computers will then "learn" what kind of land cover, what timing pattern of greening and what human densities best predict bird presence, and generate a million more simulated observations for each species: points where it is predicted to be either present or absent at different times throughout the year. The result is an animated map of bird movements. An early model of the movements of the indigo bunting (Passerina cyanea), a songbird that winters in the tropics, took five days to run on the lab's own computers. But the visualization was compelling, showing how the birds first made landfall at or near the Mississippi Delta and then used the river system to find their way to northern forests. "This shows how important the Gulf coast is early on in the migration," says Kelling.
With TeraGrid, the Cornell lab plans to marry such models to scenarios for climate change from the Intergovernmental Panel on Climate Change, in hopes of predicting migratory changes--and perhaps extinctions--for hundreds of species. The lab could theoretically do the same work on many smaller computers, perhaps relying on a "cloud" of laptops belonging to their birding volunteers, according to David Anderson, director of the Berkeley Open Infrastructure for Network Computing (BOINC) at the University of California, Berkeley. Kelling says they may--some day. "We don't have the infrastructure or the expertise now to figure out the issues of parsing all the data out," he says.
These ecological problems are a new frontier for supercomputing, says John Cobb, a principal investigator for TeraGrid at the Oak Ridge National Laboratory in Tennessee and a co-investigator for DataONE, a project sponsored by the National Science Foundation to gather and harmonize ecological and environmental data sets. With vast amounts of computing power comes the opportunity to turn the work of many amateur birders into a nuanced portrait of how species migrate. "It is a wonderful story about how they have used all those people who are enthusiastic about birdwatching and made a scientifically significant data set," says Cobb.