News & Features

eBird Science: Distribution predictions using eBird data are comparable to those from satellite transmitters

Band-tailed Pigeon by Rick Taylor/Macaulay Library

Ever since satellite technology has been small enough to put on a bird, researchers have been using transmitters to ask questions about birds that were previously unanswerable. Although some questions still can’t be answered with anything aside from satellites (e.g., precise paths of migrating birds throughout their entire annual cycle), a paper published this week in Global Ecology and Conservation shows that eBird data can be comparable to satellite data when creating species distribution models. The authors of the open-access paper “Species distribution models for a migratory bird based on citizen science and satellite tracking data” have written a great account of their research on Band-tailed Pigeons (below). Thanks to Chris Coxen, Jennifer Frey, Scott Carleton, and Dan Collins for taking the time to share their work with the eBird community.

Pennsylvania ornithologists have a great interest in this technology and the results of these inquiries.  We can learn more about the state’s Species of Greatest Conservation Need especially those that are more common off-road than along roads where most birds are counted.  eBirders help by taking to the trail and finding birds in wetlands, woods, and thickets away from the crowds.

[Team eBird note: All of the eBird data used in these analyses are available for you to use as well. Whether you’re evaluating presence-only data (as in this study) or using the valuable complete-checklist data that are used to generate models like our STEM maps, eBird data are free for you.]

When researchers want to learn more about where a poorly understood species lives on the landscape, or project how a species’ habitat niche may be altered due to climate change, they can create something called a species distribution model, or habitat suitability model. The model uses information about where a species is known to occur and then predicts where it may occur in different places or climate conditions.

Imagine a hypothetical bird that is only known from a few locations on a rugged mountain. Researchers think that the bird has suitable habitat on a few other mountains in the surrounding landscape, but they’re not sure. A species distribution model will estimate the habitat suitability of the other mountains based on how similar conditions are compared to the habitat at the locations where the bird had been documented. If the model predicts that one of the other mountains is highly suitable (and we trust our model), we can then search for the bird in those places and hopefully find more of them.

As you might guess, the accuracy of the prediction depends on the quality of the data you use in your model. Unfortunately, some species have outdated or inaccurate occurrence data. A researcher could hire field crews to do surveys to solve this dilemma, but field surveys are expensive and time consuming. Recent advances in satellite tracking technology (basically very small satellite GPS devices) also provide the option of placing tracking devices on a group of birds (as small as 100 grams!) to track their movements, but this technology comes at a very high monetary cost. What if a citizen science program like eBird could provide free occurrence and abundance records for a bird species with a small or outdated dataset? If eBird data were shown to work, more researchers could adopt this free, wide-coverage occurrence and abundance data set for use in their species distribution and habitat suitability models.

Comparison of species distribution models for Band-tailed Pigeon in New Mexico based on environmental predictors and (A) satellite tracking occurrence data or (B) eBird occurrence data. Warmer colors represent areas of higher relative suitability.

Our research evaluated how well a species distribution model based on eBird occurrence data compared to one based on satellite tracking occurrence data. Models created with satellite tracking data have been shown to be robust and reliable, so they are a good foil to test eBird data. We used Band-tailed Pigeons as our model species since our data set was part of a larger study designed to learn more about Band-tailed Pigeon habitat use in New Mexico. Band-tailed Pigeon population estimates from Breeding Bird Survey data in Colorado, Utah, Arizona, and New Mexico (what is known as the Interior population of Band-tailed Pigeons) show a 4.3% annual decline since 1968, their occurrence records are often unreliable or outdated, and we do not know much about how they use habitats in New Mexico. If our species distribution model based on eBird data can help answer questions about Band-tailed Pigeons, it suggests that eBird data can help researchers learn more about other poorly understood species of conservation concern.

We found eBird data to be a high-quality data source for habitat suitability models. Our species distribution model based on eBird data had a high overlap in habitat suitability scores with our model based on satellite tracking data, meaning both models separately predicted many of the same patches of suitable habitat areas. Management agencies could use our methods in conjunction with eBird data to create Band-tailed Pigeon species distribution models for other states, and increase our knowledge of their distribution and habitat suitability throughout their range. These models are the first step towards addressing the pressing need for more targeted surveys for Band-tailed Pigeons to better measure population trends.

We believe that our methods can also be a proof of concept solution for other avian species with conservation needs similar to the Band-tailed Pigeon. Our results also support other studies that show that eBird and other citizen science programs can be used to crowd source reliable data for scientific research. Wider adoption of citizen science data will empower members of the public to become an integral part of conservation research and lower potential funding barriers for scientists in need of robust data sets.

A huge thanks to all of the eBird users who take the time to carefully record and submit their findings, and to the eBird staff who manage and facilitate the use of this incredible resource.

If you do research that uses eBird data, and want your work featured for the eBird community as an eBird Science post, please write to us and include the words “eBird Science” in the subject.