In the last couple of years, eBird took two pretty important steps with their data: they began releasing animated occurrence maps for the conterminous U.S., and also opened the gates for downloading data from the global database. In combination, these developments give us a new glimpse of North American bird life but also serve as the firing of a starter pistol in the minds of the data hungry.
On a qualitative level, the animations represent the huge amount of birding we do on the continent and the power of citizen science. More importantly, they speak loudly to the dynamic nature of our bird populations and their reliance on resources across large geographic areas. While watching the occurrence animations it is almost possible to feel zugunruhe as birds vacate their wintering grounds, MacGyver their way through the gauntlet of stopover sites, seek fortunes in the Great North and return.
The downloadability of these data means that they now enter the ‘open source’ realm and it is now open season for creativity and innovation. The data can be summoned for kid science projects, amateur tinkering and exploration, modelling and hypothesis testing by research scientists, and for informing conservation planners and policy makers. With this great opportunity for storytelling also comes great responsibility; to use the data in a way that is both scientifically sound and ethical.
For my first stab at mining the eBird database, here are animated “observation density maps” for two migratory species: Red-winged Blackbird and Eastern Whip-poor-will. I have avoided the term ‘occurrence’ maps since the current maps are based on raw eBird observation data; no attempt was made to control for data quality (e.g., reporting biases, identification errors). The maps represent density of reported observations over time and therefore are not true bird distribution maps. Nonetheless, these methods can be used on any quality-controlled, subset of the eBird database to yield animations that could be used in websites or Powerpoint presentations and to allow us to look for spatio-temporal patterns in bird occurrence.
I’ve quarantined a pared-down version of the technical details to the end of the post. The gist of the process is to download the raw data from eBird; plot the points; create density surfaces and export these as images using a GIS (Geographic Information System) program; and stitch them together into an animated GIF file. These animated GIF files can then come to life when opened in any browser or in a Powerpoint presentation (hopefully they are working for you below!). I have also included here relatively up-to-date species range maps (so seasonal patterns in the eBird data can be compared to known species distributions) and relative abundance maps modelled from Breeding Bird Survey (BBS) data. I have also overlaid the raw eBird point data on the maps so that the seasonal patterns can be judged independently of the assumptions introduced when building the density surfaces.
A quick disclaimer: these observation maps represent the raw eBird observations. Therefore, areas of high density (creme and green colours) represent a concentration of both birds and eBirders. Furthermore, the continental-scale seasonal shifts in the data represent a combination of true changes in bird distribution and seasonal biases in eBirder behaviour. Both of these disclaimers also hold when viewing raw eBird data on the eBird site web application.
1. Red-winged Blackbird eBird observations map (2010)
(note the animations can be paused on any month in Internet Explorer using the escape key)
As spring approaches, watch as the range of observations expands into Canada and the southern Boreal forest then contracts back to the ‘year-round’ range in the winter months. In June, the current map seems to pick up high concentrations of birds (some of which might be true concentrations based on comparisons with the BBS model) at the southern edge of the Great Lakes, California, the Gulf Coast and Florida. Areas of coverage in the Yucatan peninsula (shown as gaps in the Birds of North America distribution map) appear in the eBird observations map because the density model spills over into nearby areas. However, looking at the point data, there were no eBird reports from within that area.
2. EasternWhip-poor-will eBird observations map (2010)
The Eastern Whip-poor-will was chosen to help evaluate the effectiveness of modelling a species with many fewer eBird records (because of smaller population sizes and lower detectability by observers). Core areas picked up in the eBird observation data roughly match those in the BBS model (bearing in mind that the BBS likely does not effectively estimate Eastern Whip-poor-will abundance and distribution). The raw eBird observation data from 2010 suggest that Whip-poor-wills hit Canada first near the Great Lakes in April and then spread further into the Maritimes and Prairies by May. Unlike Red-winged Blackbirds, after October, eBird reports of Eastern Whip-poor-will are scarce in Canada and the U.S. However, as noted in a Birds of the West Indies bird book i recently leafed through, vocalisation, and thereby detectability of these birds, is likely very limited at that time of year. For species with small populations and low relative detectability, like Eastern Whip-poor-wills, local data are likely more comprehensive and informative and I hope to delve into this in a future post!