Date: 1/13/18 3:53 am
From: Jon King <jonking271...>
Subject: A defense of the eBird review system
In response to Gene’s last email, I will address several points. First
there seems to frustration surrounding the filtering process in addition to
questions about how filters are made. At this point the data itself is the
best guide to making filters. I will demonstrate this point by walking
through the filter creation process for Blackpoll Warbler in the
easternmost counties. For those of you not aware of how filters work, see
here:
http://help.ebird.org/customer/portal/articles/1055676-understanding-the-ebird-review-and-data-quality-process

Here is how I make a filter. The filter I’m working on here is Douglas
County, Kansas, where I live. The first step is to peek at the bar chart I
made for northeast Kansas, noting the rough timing of arrival, departure,
and the high counts. This will inform my subsequent use of the search tool,
which basically just runs queries for me. First I query all Blackpoll
Warbler records between April 1st and May 1st. I can see that within the
past five to ten years the earliest arrival dates are clustered around
April 26th-28th with outliers on April 24th and 25th. Based upon this I set
the date limit at April 25th to accommodate sightings a bit earlier than
the cluster I was referring to. The species is still effectively rare on
April 25th, although not for long. I then repeat the process for departure
dates, finding May 30th to be a suitable departure date (see attachment 1).
There is only a single later record in eBird meaning the species is pretty
darn rare by the 30th. Moving on, I query high counts, and see that peak
counts very rarely exceed a dozen individuals per checklist ... There are
only a handful of such counts within the database. I then set the count
limit at twelve. Because Blackpoll Warbler is not expected during fall, I’m
done. For the finished product as well as filters for related warblers, see
attachment number two below. This filter I made could be considered
“tight”. Anything early or late by a few days, or even a significant count
is going to get flagged. That is the point. This is not a records committee
review list intended to capture only the most outstanding observations.
Cornell wants their filters tight and for good reason.

The process you just saw has been used to create filters for almost every
species in the far northeastern counties, and some will be remade to allow
for changes in climate that we are witnessing. The same amount of effort
has not yet been put into those filters covering other portions of the
state, especially west of the Flint Hills, as far as I know. Believe me it
takes hours and hours to do this. Managing things for the eastern counties
is time consuming enough, so refinement of those other will be someone
else's job.

As I hopefully demonstrated, eBird data itself is about the only resource
that provides enough information to make accurate filters. If I were to
repeat the above exercise on this forum, but use only literature and expert
advice as has been suggested, I would come off looking like a jerk and
insulting someone's work. Just be aware that there is usually far more
thought behind the filtering process than is generally realized, especially
in those counties that have been worked on. We are not incompetent. Take
for example Red-tailed Hawks. Are all those dark *calurus* getting flagged
in the eastern counties just because Jon King is a greenhorn?! The actual
reason for flagging these would probably fascinate some of you.

I also will point out that there are many relatively unknown patterns of
status and distribution buried within the eBird dataset. In working on the
database, I’ve noticed dozens and dozens of subtle and not-so-subtle
patterns that are not mentioned anywhere within contemporary Kansas
literature. This includes things like north-south phenological
discrepancies, east-west phenological discrepancies, spring-fall trajectory
shifts, spatial abundance discrepancies, subspecific migratory patterns,
and more. All of these things must be taken into account when creating
highly specific filters. Even local expertise fails to capture much of
this, as the big picture gained by pooling the observations of many, many
observers on a regional scale is hard for individuals to see. Future
treatises on regional bird status and distribution will be modified by
eBird, not so much the other way around.

It also seems that Gene questions the competence of the reviewers. All
Kansas reviewers are locals. Six of us have served on the records committee
before and three have not yet. If people think we’re an incompetent bunch,
then they should be concerned about the records committee given the overlap
in membership. Lastly I should mention that the records which reviewers
don’t accept can still be obtained when downloading data. There is a
checkbox on the download page to indicate that you want “unvetted” records
included with your other data. All of the descriptions for those records
which don't get accepted are still there.

In summary the filtering system is imperfect, but is improving in ways I’m
too tired to detail here. If some of you disgruntled locals want to consult
with us on the filtering system, great, let’s talk. Just be prepared to
work in a constructive manner ... There has been more than enough hostility
and disrespect in recent years from all parties involved.



Jon King
Lawrence, KS

For KSBIRD-L archives or to change your subscription options, go to
https://listserv.ksu.edu/ksbird-l.html
For KSBIRD-L guidelines go to
http://www.ksbirds.org/KSBIRD-LGuidelines.htm
To contact a listowner, send a message to
mailto:<ksbird-l-request...>
 
Join us on Facebook!