Effort-based observations enable powerful data analysis

eBird is a flexible tool, built to harness the power of the birding community in all its diversity. Everyone enjoys birds in slightly different ways, and for a variety of reasons. There are those of us who watch birds everyday in our backyards, while others consider their county or state their home stomping grounds, and bird these larger areas with equal vigor. Others still are world birders, moving from country to country in search of new birds. All are equally important to eBird. In addition to our varied geographic interests, we also record what we see in myriad ways. Some keep very refined checklists for each stop in a day's outing, recording all the birds they've seen and heard, and estimating numbers for each species. Others keep lists of birds seen at broader locations, or just note the birding highlights from their day. The beauty of eBird is that it was built to capture all of these data, and it is able to use all of these bits of information for different types of data output. 

For a while now we've been steering eBirders in the direction of submitting complete checklists of birds with associated effort information from discrete locations. The idea is to better understand bird occurrence and how it relates to habitat on the ground. The more information you give us, the better we'll be able to do that. Interestingly, the old ways of keeping 'day lists' that span large geographic areas are slowly fading out in favor of this more refined approach, and because of this we're starting to see some incredible results from data modeling. Below we outline how ALL data are valuable, but continue to stress the importance of what we can do when you provide us with more refined observations.

The Complete Checklist. When you first start using eBird you might be surprised to have to answer the following question:

"Are you submitting a complete checklist of the birds you were able to identify to the best of your ability?"

This is a critical question, and one that sets eBird apart from nearly every other tool that records single records of birds. The unit of currency at eBird is the "checklist". A "checklist" is a list of birds for a given date and location. So while a "record" is just one observation of a bird (e.g., one Pine Warbler at Kiptopeke State Park, VA, on 18 Sep 2010), a "checklist" is all of the birds associated with that birding event (e.g., everything you saw/heard at Kiptopeke on 18 Sep from 7AM-9AM while walking 1.5 miles along the trails). When you answer 'Yes' to the 'Are you reporting all species' question, it allows us to infer absence, or at least lack of detection, for a suite of species that might be possible in the region on this date. Importantly, what this means is that when we look at all of the 'complete' checklists combined across North America, we can then create distribution maps that show where the birds are (presence), but also where they were not recorded (absence). With enough data over sufficient time, we can infer true absence from regions where a species has never been recorded. We can also show where we currently lack sufficient data to say one way or the other if the species is present (see below).

Pine Warbler distribution 1900-2010.

On the map above we see the overall distribution of Pine Warbler, and the darker greens indicate areas of higher frequency. We can clearly see the northern breeding range around the Great Lakes, the area of paler green in the Midwest where they pass through on migration, and the heavily dark Southeast where they both breed and winter. Importantly, however, are the extensive areas of gray in the West and across the North. These gray areas represent 100km grid squares where we have at least 5 complete checklists, but no Pine Warblers have been reported. The tan areas, or 'blank' areas, on the map show where we lack data (e.g., across the northern Boreal forest). We can't know for sure whether Pine Warbler occurs here or not because we lack data. Each time you submit a 'complete checklist' of birds and answer 'yes' to that question, we can fill in a piece of this puzzle.

Incidental Observations. While we've tried to steer birders away from habitually entering single records of birds at a location and date in favor of complete checklists with effort, these single records are useful for some kinds of output. Moreover, they are completely appropriate when the situation calls for it. For example, when driving by a ranch several times per week, I frequently see a Golden Eagle there hunting ground-squirrels. I'm just driving past on my way to or from somewhere, and birding is not my primary purpose, but I still want to make sure the Golden Eagle makes it into eBird. In situations like this, using 'Incidental Observation' as your protocol is the right thing to do. Another perfectly appropriate scenario for using this protocol is the entry of historic data where you lack specific details about effort. You might know the date and location, but you just can't recall how long you were birding or how far you traveled. In these cases it's fine to use 'Incidental Observations'. We WILL use these observations for mapping bird occurrence (see below).

Tropical Kingbird distribution 1900-2010.

On the map above we can see the result of lots of 'Incidental Observations' mapped across much of South America. While North America is largely gray due to the volume of data there, South America is still being filled in, and much of it has little data so far (e.g. Brazil). Moreover, many of the records are from traveling North American birders who had old data to enter, and thus, lots of incidental observations. What we get is a map that shows where we have positive observations for Tropical Kingbird, but we're not able infer much about its overall range in South America because we don't have the 'negative data' from complete checklist to tell us about their actual range. All we can do is see where they are. Conversely, in North America, where we have lots of complete checklists, we can say with confidence that they occur only in southeast Arizona and south Texas, and sparingly along the Pacific Coast, with a scattering of other vagrant records.

Effort-based Observations. At eBird we use the term "Effort-based Observations" to group all of the protocol choices that require you to provide some measure of effort along with your checklist of birds. The most frequently used protocol is 'Traveling Count'. Most birders are doing this type of count when they do general birding. We ask that you provide an estimation of distance traveled and the time you spent in the field; two pieces of information which allow the analyst to do a lot more with your bird observations. But not all traveling counts are equal; excessively long ones are not as good as shorter ones for complex data analysis. As you can imagine, a checklist of 30 species recorded while traveling a distance of 1 mile over the course of 2 hours, is very different than a checklist with the same number of species recorded over a distance of 100 miles and a duration of 2 hrs (i.e., car birding). The latter is quite vague spatially, and we could do little to associate the birds you've reported with the habitat where they actually occurred. In fact, it's hard to even map these kinds of data because a bird may have been seen very far from the place where the point actually falls on the map, at the midpoint of that 100-mile long count! The shorter counts, however, are quite specific, and we can do a lot with these checklists in terms of analyzing species distribution. These effort-based, fine-scale checklists are the cream of the crop for data analysis. You've told us when (date and time), where (specific location), what kind (species list), and how many (estimated numbers), and now you're also telling us how far you went (distance traveled), and how long you spent collecting these observations (duration). When we have lots of checklists with these descriptive data, we can begin to perform very sophisticated modeling processes that associate birds with habitat, and we're able to control for many of the biases associated with observer effort (see below).