Summary

  • eBird's team designed seasonal charts to encourage complete local checklists instead of a day's few memorable sightings.
  • A completeness declaration and recorded search effort gave researchers a way to interpret an unreported species as a non-detection during a particular search.

A rare bird makes an excellent story. Another ordinary bird makes an excellent excuse to stop typing. For a website trying to learn where birds occur through the seasons, that editorial instinct was a problem. Science wanted the unglamorous entries too. Someone had to make them worth the bother.

eBird began in 2002, initially covering North America, and opened worldwide data entry by 2010. The Cornell Lab of Ornithology project let people submit observations over the Web. Its early reception was lukewarm, the team recalled in its 2011 account. Asking birders to contribute to science was only part of the bargain. The site developed tools for keeping personal observations and lists, comparing records and finding useful information for the next outing.

One feature did two jobs particularly neatly. Seasonal bar charts showed how often a species appeared on submitted checklists. A birder could use them to choose when to visit. The team explicitly described designing those charts to move contributors away from reporting two or three highlights for an entire day and towards complete lists from more precisely located stops. Better records made the charts more useful; useful charts gave contributors a reason to supply better records. The everyday bird acquired a small public-relations department.

The form asked for species, date and location, and whether the observer had reported all the species detected. Most records also supplied counts and search details such as start time, duration or distance. These fields attached a boundary to the observation: someone had looked here, for this long, in this way. Repeated visits by an ordinary backyard observer could produce valuable samples. The site had room for people whose chief distinction was turning up again.

Completeness made an omission readable. In their 2014 technical account, the authors described a research dataset restricted to lists whose observers confirmed that all identified species were reported. Unreported species could then receive zeros recording non-detection or non-identification during that search. Effort information helped analysts account for variation in detection. A short visit and a long walk carried their own context into the analysis.

There was also human work behind the neat rows. Regional filters flagged unusual species or counts for the time of year, and reviewers could contact the observer. The flag opened a conversation about evidence. Someone still had to assess the answer.

The 2011 paper illustrated the practical attraction with seasonal shorebird information for Montezuma National Wildlife Refuge, presented as a way to inform water-level management. The path from a birding trip to a useful public record ran through these mundane choices: which sightings to include, how precisely to locate them, and how much searching the list represented. The Internet could collect the reports. eBird had to give people a reason to finish them.