News & Features

Oops! Top 10 Atlas Coding Errors & New Code Reinterpretation Process

Bobolink by Ryan Brady

As the season starts to slow down and we all begin to review our checklists collected during 2017, we thought it would be informative to share with everyone the results of the data review of 2015 and 2016 data.  In general, we’re seeing lots of good information rolling in, but there are a few common pitfalls that everyone should be aware of.

 

1.). Vocalizations do not always indicate breeding

Yes, those Cedar Waxwings and Blue Jays were calling. But that’s the call they make all winter too. Those sounds do not signify that they are holding a territory, such as when a Song Sparrow is singing repeatedly from the same area.

How do you know if a vocalization warrants a code (S, S7, M) or is not codeable? If in doubt, check out our Acceptable Breeding Codes Chart.

2.). DD (Distraction display) is a rare code

Based on comments, we’re realizing some of the observations coded DD are better coded A or T.

DD codes involve either feigning injury or a physical attack on you.
– Killdeer doing its broken wing display
– A grouse flapping around trying to attract your attention
– A Red-winged Blackbird aggressively attacking you
– A Northern Harrier divebombing you

A codes involve clearly agitated birds, but don’t involve an attack or specific injury-feigning display
– A Song Sparrow scolding you
– A Red-winged Blackbird flying around scolding you​, quite mad but it falls short of directly attacking you

T codes involve a physical altercation within a species, or a physical driving away of another species (In the rare instance a predator is actually trying to raid a nest, this could be DD).
– Two American Robins in a physical fight
– A Northern Cardinal fighting its reflection in a window
– An Eastern Kingbird chasing a Red-tailed Hawk

3.). Simple Typos

Did you mean PE or P? Did you mean Red-headed Woodpecker or Red-bellied Woodpecker? After a flagged record, did you enter comments?

A quick skim of your checklist after you submit it can help with typos. As can periodically scanning the list of species and codes for your block to make sure everything got in there correctly (available through Explore a Region, and beware that it may take 24 hours to update).

4.). For wide-ranging birds, code carefully

If you hear a Red-eyed Vireo in your block in June, chances are it breeds in that block. But if you see a Turkey Vulture, maybe it’s soaring on thermals and is miles away from its nest. Birds like vultures, Bald Eagle, and especially colonial waterbirds like herons, gulls, terns, and American White Pelicans are notorious for roaming far from breeding sites even during June and July. Just seeing these species in June and July is not a slam dunk H, and codes like P, FY, and FL should be used with caution.

5.). UN (Used Nest) doesn’t mean the nest is currently being used

UN is a special code for a nest that was used earlier but is now empty. Because nests without occupants can be quite challenging to identify, use this code with caution (see the Acceptable Breeding Codes Chart for guidance). The correct code for a currently occupied nest (e.g., an Osprey sitting on a nest) is ON, Occupied Nest.

6.). The correct code for baby waterfowl is FL (Fledged Young)

We are getting a lot of FY (Feeding Young) codes with broods of Mallards and Canada Geese, but the young actually feed themselves. The proper code here is FL (Fledged Young), as these birds have left the nest.

7.). Careful with CF (Carrying Food) for larger birds

So you saw an Osprey with a fish – is that a confirmation? Not automatically. Osprey (as well as eagles, hawks, corvids and other large birds) often end up carrying their food elsewhere to consume it. If you see a bird repeatedly carrying food in a direction, then CF is more likely. With songbirds, this code is useful the majority of the time, generally songbirds consume their own food fairly quickly, but when they are foraging for nestlings or young, they end up hopping around with a morsel (or more) in their mouth for an extended period of time.

8.). Be aware that breeding timing differs among species

In general, our resident species are among the first birds to get started each year, followed by early migrants (Song Sparrow, Red-winged Blackbird, Common Grackle, Mourning Dove, Canada Goose, Eastern Phoebe) and then lastly the neotropical migrants (warblers, vireos, thrushes). But some species are surprising – Pine Siskin can have young by early May, but American Goldfinches are a very late nester, and most shouldn’t be coded before June. In most species, things are shutting down by mid August, but some species like Cedar Waxwing, Red-eyed Vireo, Northern Cardinal, Mourning Dove, and Rock Pigeon can go into September and in some cases later.

It’s a lot to keep track of, which is why we’ve prepared the Breeding Guideline Bar Chart to help guide you.

9.). Don’t forget about the C code

In some species, C (Courtship display and copulation) is actually used quite frequently, and it’s an instant Probable-level code. We’re thinking primarily about birds that advertise their territory through display rather than song, particularly: Ruffed Grouse, Wilson’s Snipe, Wild Turkey, and American Woodcock.

10.). Know the difference between Atlas eBird and Wisconsin eBird

Remember, anytime you have a breeding code, the full checklist should go into atlas eBird. Anytime you do not have a breeding code, the full checklist should go into Wisconsin eBird. Beware if you have your phone defaulting to one portal or the other, you may have to change that using the Change Portal button on the lower left when you submit a checklist. More on portals here.

 

New Process for Breeding Code Reinterpretation

So, now that we’ve identified some common errors, what are we doing about these situations? In cases where we can reasonably make an assumption that the code needs to be adjusted, (for example, you can’t S7 Cedar Waxwing or use CF for Mallard or you can’t code Black-capped Chickadee in December or Common Goldeneye in March), we are working with eBird to adjust things on the back end. What that means for right now is that we will be changing the category of the record (that is, Observed/Possible/Probable/Confirmed) but not the specific code. So if you had an Indigo Bunting chipping angrily at you and coded it DD, but we felt it was more appropriately A, we could downgrade that record to Probable. You’ll know if a record was reinterpreted because it will look like this:

Clicking the question mark will provide more information on why the code was reinterpreted. The updated category will then be the one that displays on atlas species maps and block tables. If you’d like to learn more about this process as it currently works, read more here.

Having DD and Probable is a bit counter-intuitive, because DD is a Confirmed Code and what we’ve really done is reinterpreted that behavior as an A – Agitated Probable Code and changed it from Confirmed to Probable right now.

If, after reading this article, you remember a couple checklists where you want to edit the codes yourself, feel free to do that.

Also, If you think we’ve adjusted your code in error, please email us at Atlas Central and let us know. We’re doing our best with a large dataset and although this process should improve data quality, it’s a team effort to get our records as clean as possible.

While it’s not currently feasible to contact all atlasers regarding the records that have been reinterpreted, this process is really just a midway point in development. In the next 2 years, we will be working with eBird to allow for these records to show original and reinterpreted codes (not just Possible/Probable/Confirmed categories), as well as developing a system for easily communicating to you about which if any of your records were reinterpreted.

Furthermore, please don’t feel bad if you notice a few of your breeding codes were reinterpreted. Almost everyone in this project (from Atlas Coordinators and Planners on down!) had a few records flagged.  We realize there’s lots to keep track of, and hopefully our data quality system will be catching errors that sneak into the process whatever the reason.

A few stats on code reinterpretation on the 2015 and 2016 data:

  • 6,824 records downgraded
  • 4,894 records upgraded
  • 3,923 records moved within category (e.g., CN to NB within Confirmed)
  • Net loss of 311 confirmations
  • Overall 15,641 records reinterpreted (out of 759,000 records) – 2.1% error rate

Thanks as always for your dedication to the project, and happy atlasing!