Everytime we go birding and submit an eBird checklist, we take a tiny snapshot of bird occurrence in space and time. eBird’s grand vision is to piece all these tiny snapshots together into a global tapestry of bird occurrence. This shared effort will reveal the complex relationships of birds to the environment and, as the seasons change, show how birds flow around the planet in cycles of dispersal and migration. As a result, we are thrilled to share the 2017 STEM models, which are the product of several years of refinements and improvements over the classic eBird Occurrence Maps. The new STEM (Spatio-Temporal Exploratory Model) now incorporate Canadian data in a major expansion that includes the entire Western hemisphere for the first time! These models utilize eBird data submitted by “Citizen Scientists” in a species distribution model that has been specifically developed by statisticians and researchers at the Cornell Lab of Ornithology.
eBird data are inherently uneven, with birding effort clustered around cities, well-known parks and refuges, and even at certain times of year. The new STEM models account not just for such variability in effort and time of day, but also for spatial and temporal variability. All this allows for smooth predictions across the continents and throughout the year. With the first iterations released in 2009, STEM models have since been used in numerous papers and reports including three State of the Birds Reports, most recently in 2016.
The 2017 STEM models have a large number of improvements, most notably: 1) coverage of the entire Western Hemisphere; 2) model predictions of abundance, instead of occurrence or frequency (which is why counting birds matters!); 3) predictions of bird occurrence at finer scale than before (formerly 30km, now 8km), making for more detailed maps; and 4) use of new statistical techniques to mitigate effects of over-extrapolation.
Each pixel depicted in the new models gives a specific prediction for the expected number of individuals you could expect to encounter at a given location if you went birding for 1 hour at 7am covering 1 kilometer of distance.
These models use your eBird data, so we thank you all for your effort-based complete checklists and attempts to count or estimate the species you observed. All your data submissions have value.
We are kicking off the release of these new STEM models with one of the most widespread species and one of our favorites: Barn Swallow (Hirundo rustica). As you marvel at the full life cycle of this incredible species, don’t forget to check out Bird Studies Canada’s work on this declining bird. Also, be sure to check back often to see all new STEM models here as they’re added!