In the beginning...

... there was word of mouth and, for some, publication options. For most, bird-occurrence data were unimportant and, particularly, unknown and unlooked for. With the advent of the widespread availability of telephones in houses, phone trees became established in certain cities to spread word of rare birds. Audubon societies and bird clubs spread and became sources of information on bird occurrence. Then the Internet came into being and was gradually co-opted for distributing information on the occurrence even of not-so-rare species.

Then the Cornell Laboratory of Ornithology said, "Let there be eBird," and it was... not so much good as better than anything previous. Excellent planning on the front end about some aspects of what eBird would do and how it would do it made for a reasonable stab at its first steps. The powers that be sussed that some sort of filter of incoming data would be required, but the thinking on that topic, in my estimation, was sub-par, even very poor. eBird has become much, much better since 2002 at filtering incoming data, but this blog exists because eBird has not enforced and does not enforce its filter... rules.

eBird's Help Center provides this explanation of eBird's filters:

"Most observations are flagged by automated data filters. These automated filters are the foundation of the eBird review process. They provide a first check on the species, count, location, and date of every observation submitted to eBird. Any report that exceeds expected totals for a given species at a location on a given date gets 'flagged' and requires further documentation."

That eBird, itself, describes eBird filters as "the foundation of the eBird review process" indicates the importance of eBird filters. Unfortunately, after writing those above words, eBird has done very little else to make their words anything like reality. eBird has left interpretation and implementation entirely to the vagaries and vicissitudes of the, presumably, hundreds of individual volunteer (read: unpaid) filter makers to invent their own wheels. In my experience of submitting eBird checklists in many, many filter regions of the world, most of those wheels are not perfectly round. In fact, many of them are polygons, being not at all round and virtually unable to make the eBird system travel down the information super-highway even at walking speed, much less highway speed.

A digression. From the inception of eBird in 2002, I was the primary filter maker of Colorado eBird filters for the better part of 20 years. I took Colorado from a single, statewide -- and perfectly useless -- filter region to a state of 64 counties having more than 40 filter regions. Kathy Mihm Dunning has taken over the thankless task. Yes, the task truly is thankless from an official eBird view as comments from eBird Central about Colorado filters tend more toward excoriation than exuberance... and that despite how incredibly less effective nearly all other eBird filters are. Yes, there are some tight and effective eBird filters in places other than Colorado, but they are rare, quite rare, and, as far as I am aware, very, very few states or other below-national-level entities have every subregion's filter being tight and effective.

I have written elsewhere about filters (such as here; please read this as I would prefer not to repeat all that information in this venue about how filters are made). I have spent relatively large amounts of thinking time and mental energy on eBird filters -- the whys and the hows -- and had to do that because eBird provides its reviewers/filter-makers with exceedingly little in the way of input on the nuts and bolts of filter making, leaving each such person to invent or reinvent the wheel for herself or himself or themself. This failing at the organization level greatly reduces the scientific utility of eBird data.

In the pages and entries of this blog, I will highlight individual eBird filters or individual species across eBird filters, pointing out those filters that seem to do a good job, but, more likely, how so many do so poorly at the jobs those filters are supposed to do. I will use eBird's own words to point out the failings... the oh-so-many failings, those egregious failings... of those filters.

I fully expect individual eBird reviewers/filter-makers to rail against my posts, perhaps with even eBird administration doing the same. As a first cut in anticipatory response to those complaints, I will state that making and keeping current strong eBird filters is very time-consuming. Very time-consuming. Some people have jobs, families, and other aspects of their lives (such as birding) that leave them little free time to create and maintain tight filters. Not only do tight filters require a lot of effort to create and maintain, they also create more review work, as tighter filters filter more questionable or even perfectly acceptable reports into the review process than looser, less precise filters do.

To close, I believe that eBird should hire a few people whose sole task for eBird is making filters and revamping existing filters to make them more effective at a filter's task, that of providing a strong, first-pass review of data coming into the eBird dataset. The eBird administration knows that to be true, but does not make it happen for reasons that are perhaps legion, but are also not communicated. However, this suggests, at least to me, that eBird is not truly interested in having effective filters given that eBird does not enforce on filter makers their own suggestions/rules.