Lubbock County, Texas, USA

Since it was included as an example of eBird filters in eBird's Help Center, I am starting with a critique of this eBird region's filter. I happen to have experience submitting data to eBird in the county/region, so am personally aware of some of this filter's inadequacies, and I will begin very high up in the taxonomy with Cackling Goose.

CAVEAT: I have no idea if what is presented in the aforementioned Help Center article reflects the current Lubbock County filter, because I have no idea whether or when that filter has been altered since eBird captured a snippet of the filter to use in that explanatory page. However, that ignorance does not affect my arguments below about filters and filter limits.

Cackling Goose is an eBird problem child for two reasons, the first being that many eBirders are unaware of the potential for the species to occur nearly anywhere in the United States, and, second, even of those who are aware of the above, many have poor understanding of how to distinguish the species from Canada Goose and its many subspecies of greatly variable size. While many online sources provide some measure of instruction in distinguishing these two species, most are too brief for such a large and difficult problem. David Sibley provided a good more-than-just-overview. Throw in various cross-species hybrids and intraspecies Cackling Goose intergrades, and the problem of correctly discerning the referable species of a single white-cheeked goose can cause confusion and uncertainty within the careful eBirder, even one who can distinguish between large subpecies of Cackling Goose and small subspecies of Canada Goose. The relatively poor treatment of Cackling Goose on many eBird filters does not at all help the situation. The Lubbock County filter makes things even worse.

If one initiates an eBird checklist for a location in Lubbock County on any day of the year, Cackling Goose will be listed as an option with a non-zero filter limit, that is, the filter will allow some number of individuals to be reported. As I write this and have open an eBird checklist from the county, I can play with the number entered and find that the filter limit for July 15 is 10. That is, I can report as many as ten without having to provide details of what would actually be an extraordinary report, even if I were reporting only a single bird!

By asking the eBird database for a map of reported Cackling Goose occurrence in July, one can get a hint about the problem of loose, overly generous eBird filters. If one gets more curious and asks eBird for the occurrence graph of Cackling Goose in Lubbock County, one gets much more than just a hint. Oh, because of the very high numbers present in the county in late fall through early spring, the summer numbers reported in the county appear at first glance to be zero. By selecting the Average Count tab and putting the cursor within the graph in July, one finds the four eBird weekly average counts (weeks starting on the 1st, 8th, 15th, and 22nd) in the county to be 2.27, 2.94, 3.08, 3.67 birds per checklist of checklists in the county reporting the species at the time of year. As far as I can easily determine, these values are higher than anywhere else in the world in July... except for areas within the arctic breeding range of the species.

As a double-check, we can compare Lubbock County's Canada Goose data with those of Cackling Goose data there; the result is eye-opening, perhaps downright scary. The four weekly average counts of Canada Goose are 7.38, 8.9, 10.92, and 15.4. For those with less math facility, those numbers indicate that, in Lubbock County, Canada Goose is more common in July there than is Cackling Goose by factors of only 3.03 to 4.20. More simply, using those data, if one runs across ten white-cheeked geese in July in Lubbock County, three to four of them will be Cackling Geese.

As a triple-check, one can compare the Lubbock "data" with those from other locations on the plains that do or probably host relatively large numbers of Cackling Geese in winter or during migration. Taylor County, Texas, which hosts the city of Abilene, which is just 167 road miles to the southeast of the city of Lubbock, has, seemingly, the opposite problem: Virtually no eBirders report the species at any time of year, much less in summer. Given the relatively large numbers of Cackling Geese reported from Lubbock County in winter (max average count of 800 in the third week of January), I find it extremely unlikely that essentially zero occur in the Abilene area. I may have to go wander around Taylor County this winter.

Continuing the triple-check, the Jefferson County, Colorado eBird data provide the sort of seasonal Cackling Goose picture I understand, with a peak average count of 178 during the third week of December and with the four weekly average counts in July being, 0, 0, 0, and 1. Given that the county is the most highly eBirded county in Colorado, eBird provides a good synopsis of occurrence patterns of large, easily visible bird species. (For comparison's sake, Jefferson County's total of ~237,000 eBird checklists compares favorably with the state of Wyoming's ~265,000 and Utah's ~652,000 checklists all time.) Finally, in our triple-check, Sedgwick County, Kansas (home of Wichita) has the four July weekly average counts of Cackling Goose of 0, 0, 0, and 0.

That's Cackling Goose. Now, onto the portion of the Lubbock County filter presented in eBird's Help Center regarding filters (see link in the first sentence).

The first species shown is Chihuahuan Raven, with a year-round limit of 10. Again, such a filter limit allows eBirders to enter as many as ten onto a checklist from a location within the county without the entry being subject to eBird review. If that filter limit seems like a shot in the dark or a by-the-seat-of-your-pants guesstimate, you are correct, as only seven of the 52 weeks of Lubbock County provide for a high count of more than 6 and only 19 high counts were greater than 4. One week does have a high count of 1475! More importantly, the high count during the period of May through July is of only 3, Given that the lowest weekly number of checklists reported from the county in that three-month period is 286 and the max is 754, it seems obvious to me that running across ten Chihuahuan Ravens during the period of one checklist is exceedingly unlikely and any such report ought to be reviewed, particularly given the difficulty some observers have at distinguishing between crows and ravens... and Great-tailed Grackles.

Let us move down the taxonomy to Horned Lark, for which the Lubbock County filter distinguishes five seasons with differing filter limits: 250 (1 Jan through 29 Feb), 100 (1-31 Mar), 501 (1 Apr through 31 Aug), 100 (1 Sep through 31 Oct), and returning to 250 after October. Looking at the county's high counts, we find only one week -- the last week of December -- in which the all-time high count of Horned Lark exceeds 174. During no week other than that last week of December does the county's high count of Horned Lark exceed the relevant filter limit, although one week -- last week of June -- matches the filter limit. The take-home message is that the Lubbock County Horned Lark filter... does virtually no filtering. The weekly high counts in March in which the filter limit is 100 are 50, 6, 10, and 15.

Farther down on the snippet of the Lubbock County filter is Tree Swallow, which has filter-period start dates including those of 4 Mar and 19 Apr, and this illustrates one of my greatest peeves with eBird filters. While the 4 March date of the opening of the seasonal period in which the Tree Swallow filter has non-zero limits may simply indicate the earliest-ever date for the county, the Tree Swallow bar chart of occurrence there proves that thought wrong: There are accepted data from as early as the second week of February!

Perhaps, then, the 4 March date indicates when species generally arrives in the county. While technically not provably incorrect, Tree Swallow frequency on Lubbock County checklists is 0.77% during the last week of February, but only 0.3% in the first week of March and 0.7% in the second week of March. Remember all this, as it is relevant to my primary complaint about filter periods in eBird filters.

The 19 April start date of the third period in the Tree Swallow filter seems perfectly arbitrary, possibly another of those seat-of-the-pants WAGs. Having formerly had eBird review permission for all of the US and Canada, I've looked at a lot of filters. Considering that I'm probably far from the smartest reviewer ever, why is it that so few reviewers/filter-makers have seemed to set filter periods using existing eBird data (both timing and abundance) and, particularly, how those data are presented in eBird graphs: by week, with the four monthly weeks beginning on the 1st, 8th, 15th, and 22nd. Yes, that fourth week is quite variable in length (7-10 days), but those work very well at provding a focused view of bird occurrence in most eBird filter regions. In my opinion, most filter periods should start on one of those four dates in whatever month is being considered. Yes, there are migrant species that arrive in the last few days at the end of a month (such as MacGillivray's Warbler in low-elevation portions of Colorado in April), but filter-makers should be familiar enough with occurrence parameters within their region(s) to know those sorts of exceptions.

I here answer the question I posed in the previous paragraph: Laziness... or limited free time. However, were I the project lead for eBird, I would greatly prefer having local reviewers/filter-makers not exhibiting either of those features.

While I believe that having filter periods of lengths supported by existing eBird data, even more firmly, I believe that filter limits should be a balance of the usual sorts of maxima reported and the infrequent higher counts reported. That is, filter limits should NOT be simply the highest tallies recorded for the filter period, else why have a filter at all?