The Bird Climate

Earlier this week, The New York Times reported on eBird, a network of bird observers using smartphones to collaborate on the vast project of making a global picture of bird populations. Launched in 2002 by the venerable Cornell Lab of Ornithology, the network has already compiled nearly 150 million reports of bird sightings, and the amount of data it receives each year continues to grow. Poignantly, eBird is also promoted as a way to prevent the diligent observations of disaggregated bird-watchers from being lost -- to science and, by extension, to eternity. 

Dr. John W. Fitzpatrick, director of the Cornell Lab of Ornithology, comments on this in the article:  

“People for generations have been accumulating an enormous amount of information about where birds are and have been.... Then it got burned when they died.”       

The eBird network saves this information from the fire, so to speak, by converting it into data - accumulated, centralized, and brought into sensible communion with other data.

The dynamics of this data, the constant addition of new information about bird sightings, and the scope of the eBird database distinguish it from previous efforts, such as the Audubon Christmas Day Bird Count, which also organized amateur birders, bird lovers, and pro ornithologists (initially in the Northeastern US, later across the North America) for a one-day extravaganza of bird watching, identifying, and tallying. In contrast to this "static" one-day count of these moving objects, what eBird makes possible is a conception of birds as a phenomenon like climate -- global, interconnected, dynamic. If the Audubon Christmas Bird Count is the local bird weather report in various locations on a particular day of the year, eBird is the global bird climate: the patterns and moving fronts, with concomitant capacity to make predictions about future local bird weather. The scientists who use the program even call their records of particular species a "heatmap." 

The birders who participate in eBird aren't just ordinary birders, they are -- in eBird's words -- "biological sensors," nodes in a technosocial network to produce knowledge of the bird climate.

But as in any case where bodies and machines come together, there are ticklish issues at the interface. Though humans may be the best bird detectors, they lack some of the qualities of machine parts: consistency, reliability, regularity, standardization. And so the biological sensors' information, entered via the eBird smartphone app, has to filter through other humans - the Cornell Lab of Ornithology - to be sanctified as data. The information has to pass through the experts. These experts may also avail themselves of machines: the Times reports that eBird's creators are trying to make up for the variations among its biological sensors by using "machine learning" to "train" their program to distinguish signal from noise, to flag and discredit false or misreported or misidentified sightings. And they are also curbing bad data the old-fashioned way: by sending scientists out to refine the capacities of the biological sensors, training non-scientist eBird users to make the correct calls.

One thing the article gets a bit wrong: the Times article claims that prior to eBird, one-day counts were the only source of information about bird populations. The archetypal example of this is the Audubon Christmas Day Bird Count, which began in 1900. I'd also argue that bird banding, which was first used as a scientific method of tracking birds around the same time the Bird Count began, is another major source of information about bird populations, migration, and behavior. It's no coincidence that both the bird count and bird banding appeared at a similar time. If bird migration had long been a phenomenon of scientific interest, at the turn of the twentieth century, organized networks of ornithological observers (proto-eBird) affiliated with institutions like natural history museums, national governments, or conservation groups, made viable the vast data collection project entailed by the study of migration.

Illustration of birds dead on the pavilion below the Statue of Liberty's torch, from  Duluth Daily News , November 8, 1887

Illustration of birds dead on the pavilion below the Statue of Liberty's torch, from Duluth Daily News, November 8, 1887

There's yet another, lesser-known, source of information about migration that was also used at this time: birds that collided with human-built structures. The Statue of Liberty's electric torch first blazed in 1887; the statue of William Penn which crowned Philadelphia's city hall (briefly, the world's tallest building) was floodlit in 1898. Under certain weather conditions -- drizzle, low cloud cover -- hundreds of migratory birds were killed on certain nights in collisions with these or similar structures, other monumental electric-lit structures in the still largely gas-lit city. As one 19th-century article describing the avian casualties at the Statue of Liberty put it, these "victims of liberty and their love of light."

Moreover, these were not urban birds - sparrows and pigeons - they were migratory birds passing along ancestral flyways, forest dwellers and waterbirds rarely seen in the city's vicinity.

What happened to the bodies of these birds? 

At a time when feathers for ladies' hats were a hot commodity, these bodies could have been plundered for their valuable plumes. Colonel Augustus Tassin, who was in charge of the Statue of Liberty grounds, did not allow this to occur. He told a newspaper reporter in 1887:   

“I have heretofore received many letters from all sorts of people offering to buy the birds which were killed in this way. But I believed they were public property, and that I had no right to dispose of them.... When I have collected about 200 specimens, I send them to the Washington National Museum, the Smithsonian Institution, and other scientific institutions, where I know that they are wanted.”

Indeed, the Smithsonian's 1888 Annual Report records the receipt of 260 birds of 40 species "in the flesh" from Tassin, recognized as one of the "more important accessions during the year." Government regulations required Tassin to record data about avian fatalities at the Statue of Liberty, which was technically a lighthouse and thus subject to this requirement. But the practice of scientific collecting at bird collision sites was adopted at other late 19th- and early 20th-century urban civic sites that saw similar mass fatalities.

The tale thus becomes a sort of redemption narrative, a conversion of meaningless death to meaningful data – and reclaiming the specimens as public, scientific property rather than private commodities.  

Further, the data produced by bird collisions had certain advantages over information from bird sightings during migration. What you had were the real bodies of birds, material specimens. This allowed ornithologists to make note of things that a sighting cannot provide a clue to: the bird's final meal, its sex, its approximate age, its weight. At the turn of the 20th century -- a time when the issue of "scientific collecting" (killing birds for research) was drawing sharp scrutiny and criticism from emergent conservationist groups like the Audubon Society -- bird collisions provided specimens that illuminated the phenomenon of migration while evading the question of whether killing wildlife was justifiable on scientific grounds.

This practice continues to this day. Birds that die after colliding with buildings in New York, in Chicago, in Philadelphia, and other urban areas are collected by bird collision monitors, bagged and tagged and incorporated into natural history collections and also used to raise awareness about the vast fatal scope of glass and architecture on bird life in, above, and around cities and other places where people live and build shiny or disorienting things. (Not every collision is fatal; many of these groups also save and rehabilitate wounded birds.)

Which brings me back to Dr. Fitzgerald's quote at the beginning of this post, that the collection of data is a way to prevent loss, to stave off the fire of oblivion.

Bird specimens lead productive afterlives in natural history collections, and continue to yield information about population genetics, historical ecology, behavior, and physiology, among other things. But making a bird a specimen entails loss - things that are discarded in the process of bringing the bird's body into conformity with the other bodies in the regimented drawers in the back rooms of natural history museums. Likewise, eBird certainly allows the birdwatcher to give her or his observations a rich and productive afterlife. But that shouldn't stop us from asking: what might be lost here? What does not pass into the eBird data set? And does that absence matter?