Friday, March 18, 2016

eBird checklists and miles traveled

One of the tasks that eBird reviewers find difficult is dealing with checklists that involve long, traveling counts.  Such checklists often stand out because they tend to rack up relatively large numbers of individual species, thus tickling the relevant filter and getting particular species entries flagged.  However, a large percentage of such checklists have no flags, thus do not see review until and unless a reviewer goes searching for anomalous eBird locations of habitat-restricted species, such as, say, White-tailed Ptarmigan.  In fact, these habitat- and range-restricted species are the primary argument for having checklists with as precise a location as possible, else the resultant eBird maps show occurrence in areas or habitats in which the species does not occur.

Unfortunately, the most comprehensive treatment of the subject is not that comprehensive and does not provide any hard-and-fast ground rules.  Granted, hard-and-fast rules when dealing with biology are nearly impossible, but eBird does care, and care strongly, about checklist locations.  In fact, I'm sure that if it were possible, eBird would prefer a checklist and associated precise location for every individual bird.  Of course, that is not going to happen, at least, not anytime soon.  However impossible the every-bird-is-its-own-checklist pipe dream might be, there are definite and very strong advantages to eBird of having checklists cover as little distance as we can get from eBirders.

The best guidance that I have found in the eBird 'Help' tab is the "Traveling Counts" section of a general-purpose essay on making checklists more useful:

Traveling counts have proven to be the most effective type of observation for modeling bird populations at large scales.  By doing these counts, birders often detect a good proportion of the birds in a given habitat.  It is critical, however, that your traveling counts not be too long.  Our analysts are able to effectively use traveling counts that are ≤5 miles [emphasis added].  Most birding that is conducted on foot easily falls within this window, but traveling counts by car can often be longer.  Please consider breaking up your long traveling counts into shorter-distance ones. It's best if these shorter counts are in a relatively consistent habitat, or do not pass through habitats that are too different. For example, a logical point to break a longer route into segments would be a transition between forest and farmland, as the birds found in these two habitat types are vastly different. Doing so would make information associated with each location — such as vegetation information from satellite images — more informative.  Plot your location at the center of the area traveled, not at the start point or end point.  It's okay to stop and spend time searching flocks of birds more thoroughly on traveling counts, as we are not assuming that you are traveling at a constant speed.  You're birding after all!
I here note that eBird Central has repeatedly stressed that this "five-mile limit" should NOT be considered cause, in and of itself, for checklist invalidation, that the limit is simply one that is very useful in particular types of analyses, such as the STEM maps of migration of individual species.  Checklists covering trips of a reasonable longer distance are certainly permitted in eBird and most such provide perfectly valid and useful data.  The problem that eBird reviewers face is where to draw the line on "reasonable."

Certainly, any eBirder should recognize that a checklist covering 650 miles on I-80 and crossing a state boundary (possibly, multiple state boundaries) and who knows now many dozens of county boundaries, does not provide data useful to anyone other than the observer.  In fact, such would provide misleading information, as almost no matter where such a checklist were plotted, the list of birds could well include species not regularly found at or near the plotted location.  But where lies the "reasonable" cutoff between 650 miles and five miles?  Unfortunately, this illusory line is dependent upon where you are and what sort of areas are covered by the checklist.

In much of the flatlands of Wyoming and Colorado, a checklist covering 15 miles, even 20 miles, might include just one or two habitats (e. g., native grassland and winter-wheat fields) that share a large proportion of bird species present in the area, and thus provide reasonable eBird data.  However, in places like southwestern Huerfano County, Colorado, traveling on the road from Highway 12 toward Trinchera Peak -- the straight-line distance of which is just over 4 miles (though the road distance is a fair bit longer) -- one's elevation changes from roughly 9250 feet to over 12,100 feet.  This road takes one from about the upper limit of elevational range of species like Pygmy Nuthatch and Black-headed Grosbeak and the lower elevational limit of Gray Jay and Pine Grosbeak to the above-timberline habitat of White-tailed Ptarmigan and American Pipit.  Thus, any checklist covering the whole road will, perforce, require plotting some of the birds encountered in habitat different from that in which they typically occur.

The above brings us back to the aforementioned eBird guideline about traveling distance and its emphasis on habitat.  Though not stated precisely in the guideline, eBird is very much interested in checklist locations being fairly specific to the habitats of the species included on the checklist.  Thus, in areas of the two-state region in which many habitats meet and mingle -- particularly obviously differing habitats, such as grassland and woodland or conifer forest and alpine tundra, the more precisely that locations are plotted and the shorter distances that are covered by individual checklists the better.  With the eBird application (app) for smart phones (about which, please read the best-practices essay), this becomes quite easy to manage.  Granted, cell coverage in much of the two states leaves much to be desired, but there is a simple work-around when planning a trip to an area of poor cell coverage.  One can modify this technique when already out and unexpectedly (or otherwise) run into poor cell reception, by using an existing checklist in the app on which to base the new one.  It is important to remember in such cases that county boundaries are often filter-region boundaries and species that are expected in some counties (thus, not flagged by occurrence, alone) may be unusual in a neighboring county, which may cause some slight disconnects between what species one sees and the "allowable" list of species on the current active checklist on the app.

Most eBirders understand the need to separate sightings in one county from those in another into different checklists.  While this is less of a problem in Wyoming than it is in Colorado (mostly due to the many fewer counties in Wyoming), both states host birding locations that cross county borders, with Yellowstone N. P. being the mother of all examples in the two-state region (heck, the Park even straddles multiple state borders!).  In Colorado, particular problem children birding sites include Chatfield S. P., Jumbo Reservoir, and Comanche N. G.  However, the single most-recalcitrant problem child is the area around Denver International Airport (DIA).

The DIA problem was created when the airport was first proposed, because Denver County was allowed to annex a sizable chunk of what was then Adams County in order to build the airport and the access road (Peña Boulevard). This action created a long, snaking corridor of Denver cutting through Adams.  Were it not for Denver County's small size and its near-complete lack of native habitat, the DIA problem would not be a problem.  However, with birders' interest in county listing, this long snake of a Denver extension travels through some excellent habitat otherwise lacking in the county and providing access to species that have been extirpated from the original county boundary... such as Burrowing Owl.  The whole eBird-review problem here is based on the fact that this corridor of Denver is so narrow, just 1.3 miles wide at its narrowest, thus it is nearly impossible on a birding jaunt of the area to not cross into and out of Adams County, sometimes multiple times.  This all means that most (if not nearly all) eBird checklists submitted from the DIA environs cover an area claimed by multiple counties, something that is not "permissible" in eBird.

In summary, I provide what I think are some reasonable guidelines on when to end the current eBird checklist and start a new one.
  • When a state boundary is crossed
  • When a county boundary is crossed
  • When crossing an obvious boundary between open habitat and wooded or forested habitat
  • When arriving at a birding location (particularly one that is an eBird hotspot) that has well defined borders and that is managed by a particular government agency or non-governmental organization (e. g., a state park, a national wildlife refuge, a private preserve)
  • When leaving a birding location (particularly one that is an eBird hotspot) that has well defined borders and that is managed by a particular government agency or non-governmental organization (e. g., a state park, a national wildlife refuge, a private preserve)
  • When leaving the "understood" area of an eBird hotspot, even if that hotspot lacks well defined borders (e. g., "Yellowstone NP--Midway Geyser Basin" or "Rocky Mountain NP--Endovalley")
  • When accruing 15 miles of distance in relatively homogeneous habitat in open country (such as in sage shrubsteppe, grassland, agriculture)
  • When accruing 5 miles of distance in relatively homogeneous forested habitat without significant (750 feet) elevational change
  • When accruing 2 miles of distance when encountering relatively significant elevational change (>750 feet, 1500 feet)
  • When accruing 1 mile of distance when encountering significant elevational change (>1500 feet)
  • When accruing two hours of observation (one hour would be even better)

1 comment:

  1. Tony, I bird Hanover Road, a hot spot in El Paso County, CO. A quick read of the checklists for this spot reveals distances traveled ranging from stationary to 15-miles or longer. While using the 15-mile entry always makes me feel a little off the mark, I have used it even though most of my Hanover Road checklists represent a collection of habitat-based shorter routes and stationary points. I suppose Hanover Road could be considered a “relatively homogeneous” stretch of habitat, but I don't think it is.

    Regardless of the distance entered on my checklists, I almost always use the hotspot “Hanover Road, El Paso County, Colorado” as the location. This means I generally enter three or four checklists for a single morning of birding along Hanover Road, each with a different distance traveled. For example, on April 12, 2016, I entered two checklists for the Hanover Road hotspot, one for 3.0 miles and the second for 5.0 miles.

    Perhaps a potential data inconsistency with using the Hanover Road hot spot is the map location. William's Creek is generally one of my stops on Hanover Road, but William's Creek is 5.4-miles west of the plotted location for Hanover Road. The wooded area near the intersection of Old Pueblo Road is even farther west.

    After reading your blog, I'm thinking I should ditch the hotspot and start using personal sites for all of my Hanover Road checklists. Doing so will shorten the distances traveled and improve accuracy of the location birded. However, I'm not certain that ditching the hotspot is the best way to go because the hot spot is a quick way for a traveling birder to evaluate a set of sites and decide where to bird.

    What do you think? Do you recommend using a set of personal locations for Hanover Road, or should I use the Hanover Road hotspot and vary the distance traveled/times for each of the segments and points birded? richard bunn