I propose that we use the new bioregion lines to create a set of 1-6 custom
filters within each county. Hopefully most counties would have 2 or 3 custom
filters. We would work together to create new bioregion filters which could be
applied to each custom filter within a county. County reviewers would maintain
these filters and tweak them as needed to reflect unique bird distributions for
that county. The Colombia and Coastal bioregions would also be broken into
individual filters within county boundaries, but the Pelagic bioregion would be
its own filter without the county boundaries.
I propose that if a county reviewer, such as Matt in Douglas County, feels that
multiple filters would be problematic, we just leave that county’s existing
filter intact.
Joe Blowers
On Mar 15, 2016, at 1:14 PM, Ian Davies <id99@xxxxxxxxxxx> wrote:
Hi everyone,
Just wanted to jump in here quickly, since I’ve been involved with the
delineation and implementation of a lot of these custom filter regions
throughout eBird in the past several months. I think that everyone here is
working towards the same goal, but that there are certain pieces that could
be clarified to make the conversations go more smoothly.
First off, a big thanks to Doug for taking the initiative on getting these up
and running, and to everyone else who has chimed in with their thoughts and
concerns. It is really great to see a group of people so engaged in eBird
data quality.
The definition of a custom filter region is only that it contains regions
that differ from the current geopolitical region boundaries. Beyond this, it
can be whatever the review team wants it to be. We can create anything from
large-scale bioregions that cover wide swaths of the landscape, to a
different filter for each street in a city. All we need for implementation is
a GIS shapefile that contains all of these custom filter regions, each with a
name and associated filter to be assigned to. The difficult part, and the
aspect of this that has been wrestled with in this forum, is how to strike
the balance between too many filters, where there are just too many different
filters to edit and maintain, and having too few filters, where data quality
for geographically restricted species suffers, or requires manual
intervention (e.g., someone searching on the Species Map to curate records—a
great thing to do, but good to make as unnecessary as possible).
One of the areas where there have been questions is how these new filters
will interact with the older geopolitical filter regions. The new filters act
like a brand new layer of data quality, sitting on top of all of the current
geopolitical filters. This means that if there is a custom filter region that
is implemented for an area, the county filters will cease to be used for
sightings in that area. I think of it like dropping a sighting pin onto a
surface. The surface is the current geopolitical boundaries. The custom
filter regions sit on top of this surface. If there is a custom region that
has been implemented, the sighting hits the custom region, never reaching the
geopolitical. If there is no custom region, it just hits the geopolitical
region like normal. There do not have to be custom filter regions in the
entire state—if they aren’t created, the old (underlying) geopolitical
filters will be used instead.
With this in mind, we can create custom filter subdivisions for certain
counties, leaving others as they are if they don’t need it. We can also have
location-specific cutouts (e.g., Malheur), that allow us to have more
stringent filters for a surrounding region. Or we can have broader custom
regions that cover multiple counties; allowing data quality in multiple
counties to be adjusted by editing just a single filter—and not having that
filter only be beholden to county-line boundaries.
By splitting up the region into areas that make sense for bird distribution,
instead of fairly arbitrary geopolitical regions, we can certainly improve
the automated data quality processes, and also the eBird user experience.
Remember also that every Likely list of birds (whether on the web application
or eBird Mobile) uses the filters to deliver the “likely” list of birds. This
means that more fine-scale filters means better-tuned lists for eBird data
entry.
In summary, we view these filters as an opportunity to further fine-tune the
data-quality process in eBird; allowing for more intelligent filters that
take into account the regions that make sense for bird distributions. Since
most of us tend to think about birds in terms of a geopolitical landscape
(myself included!), it often makes sense to split these along county lines.
As long as everyone is alright with maintaining multiple filters within their
counties of review, then that seems to be the way that best resonates with
people. For some people that review wider regions (e.g., Northeast Oregon?)
we could implement a couple regions that span county borders; potentially
even lowering the number of filters that need to be maintained.
I hope that this helps clarify the options and gets everyone on the same
page. Please let me know if you have any questions, and I look forward to
seeing the final set of new regions!
Best,
Ian
--
Ian Davies
eBird Project Coordinator
Cornell Lab of Ornithology
http://ebird.org/content/ebird/ ;<http://ebird.org/content/ebird/>
On Mar 15, 2016, at 3:34 PM, Noah Strycker <birdboy@xxxxxxxxx
<mailto:birdboy@xxxxxxxxx>> wrote:
Hi Doug et al. -
Oh! I'm honestly not sure if that will help me much in Harney and Malheur
counties. The new northeast Oregon biofilter includes everything up to
Wallowa County and west to Bend, and the southeast one is also massive. That
means replacing a smaller, very well-known filter (Harney County) with two
much larger regions that cover about half of the state, in hopes of gaining
better resolution.
Here are some birds that do not occur regularly in northern Harney County
that would now show up as common defaults in that area (this is the "error"
we are introducing):
Spruce Grouse
Gray Partridge
Mountain Quail
Great Gray Owl
Barred Owl
Three-toed Woodpecker
Chestnut-backed Chickadee
Veery
Gray Catbird
American Redstart
Grasshopper Sparrow
Tricolored Blackbird
Pine Grosbeak
Purple Finch
Gray-crowned Rosy-Finch
I'm sure there are others, but that's a start. In many ways, northern Harney
County has more in common with the rest of Harney County than it does with
far northeast Oregon, even though it slices a little bit into the Blue
Mountains. It really doesn't matter to me - we can keep the southeast region
as it is, or switch to the new bioregional filters - because I don't think
it will make much difference in data quality either way (it will just
rearrange where the errors come in).
Might we consider an alternative? I think we could use this new bioregion
map to simply subdivide some of Oregon's counties into smaller regions, on a
county-by-county basis. That way, we could use the existing county filters
as a starting point for the subfilters and each reviewer could tweak them
from there as needed. I think that would be less work; it would incorporate
bioregions while building on what we already have (thus a true hybrid
approach); and it would capture much better resolution. When this process
started, that's what I was envisioning ;)
Thanks again for all your work on this - hopefully we can catch more of
those list-building submissions. I saw a Brandt's Cormorant reported from
Fern Ridge the other day! Sorry to have missed the meet-up in Corvallis (I
was at the Klamath bird festival), but sounds like it was productive.
Good birding,
Noah
On Tue, Mar 15, 2016 at 9:44 AM, W. Douglas Robinson
<w.douglas.robinson@xxxxxxxxx <mailto:w.douglas.robinson@xxxxxxxxx>> wrote:
Noah,
County filters only function if there is an unexpected/unnoticed gap in the
biofilter. They are a "back-up" in that sense.
So, you should be thinking about this as biofilters doing all the work. Then
tripped records get sent to your county-based review queue so you can handle
them.
Doug
On Mar 15, 2016, at 8:54 AM, Noah Strycker wrote:
Hi all,
Thanks for the continuing discussion! I agree that these bioregion filters
will help in some areas, but I'm also not quite clear yet on how they will
be implemented.
The main problem we are addressing is that some birds are found only in
parts of counties. So, by subdividing the counties by bioregion, we can
catch more erroneous reports.
My question is: Will the existing county-wide filters and new bioregion
filters function as equals, or will one take precedence over the other?
In other words...
1. If Species A is rare in the county AND in the bioregion, then it should
get flagged.
2. If Species A is rare in the county BUT NOT in the bioregion, it should
still get flagged.
3. If Species A is NOT rare in the county but IS rare in the bioregion,
then it should get flagged.
4. If Species A is NOT rare in the county and NOT rare in the bioregion,
then it should not get flagged.
We are mostly talking about #3 above, right? I just want to make sure that
the new bioregions aren't taking precedence over the county-wide filters,
so that we can avoid introducing any #2-type mistakes. Any report that
trips EITHER filter should get flagged for review, not just the bioregional
filters.
That said, I don't think we need to create new filters at all. It seems to
me that the easiest solution would be to subdivide *each county* along
these bioregional lines, then initially set the new subfilters to be
exactly the same as the existing county-wide filters (which are already
very good). Then I can add the tweaks I want for each subregion in my
review area - like setting White-headed Woodpecker to be rare in most of
Harney County - without any needless wholesale changes. The idea is to gain
more resolution within counties, not to lose resolution by having sections
of each county become part of some larger bioregion.
Good birding,
Noah
PS - the .kmz map divisions in northern Harney County and Malheur County
look fine to me. I don't think we need any "special areas" around Malheur
or Steens Mountain. In Lake County, the isolated pine forest on Hart
Mountain (which has White-headed Woodpeckers, Flammulated Owls, and other
forest birds) is mapped within the southeast desert - not necessarily an
issue, but another good example of why we need subfilters at the county
level ;)
On Mon, Mar 14, 2016 at 7:10 PM, Jeff Harding <jeffharding@xxxxxxxxxxxxxx
<mailto:jeffharding@xxxxxxxxxxxxxx>> wrote:
Doug, and all,
I think the regions are fine, as they cross Linn and Marion Counties. Linn
and Marion counties would not be well served by having separate filters from
the surrounding counties. I'm sure we can work out procedures for handling
cross-county filters. Ankeny could have its own treatment, but the
differences aren't so very great, given developments at Duckflat Road,
Diamond Hill, and the like.
Thanks for moving along with the project,
Jeff
-----Original Message-----
From: orebird-bounce@xxxxxxxxxxxxx <mailto:orebird-bounce@xxxxxxxxxxxxx>
[mailto:orebird-bounce@xxxxxxxxxxxxx ;<mailto:orebird-bounce@xxxxxxxxxxxxx>]
On
Behalf Of W. Douglas Robinson
Sent: Monday, March 14, 2016 4:07 PM
To: orebird@xxxxxxxxxxxxx <mailto:orebird@xxxxxxxxxxxxx>
Subject: [orebird] More on the filters and next steps
Hi all,
I spoke today with Ian Davies at eBird about the filters. The conversation
really helped me so I hope I can communicate clearly some of the key points.
First, there are two major types of layers in eBird, the geopolitical layer
with counties, states, countries, etc, that covers the whole globe; and, in
some places, a custom filter layer associated with biogeographic
characteristics. That second layer overlays the geopolitical one. It is the
custom bioregion filter that we have been discussing for Oregon. So, current
eBird review in Oregon obviously only functions with the geopolitically
based bird data filters we have created and maintain.
What seems to concern some of us is how the bioregional filters would work
and if we implement them, do we gain enough to make the effort worthwhile.
We have 3 options.
One, leave everything alone and run Oregon data screening based on
county-level filters. The advantage is familiarity as this is our status
quo. The major drawback here, which we discussed during the meeting in
Corvallis, is that we lose track of important major habitat differences
within counties and some erroneous bird records can easily escape our notice
(i.e., those records do not trip county filters because the filters are set
too "coarsely" to accommodate the wide variety of bird records that may
occur within a diverse county, ocean birds get mapped on land, mountain
birds get mapped in a valley or in the desert, etc).
Two, shift to a completely habitat-based filter system (bioregions) and go
away from county-level filtering altogether. This is happening in diverse
tropical countries where small shifts in elevation and even aspect can
radically alter expected species and their numbers. If Oregon went this way,
we could still assign reviews based on counties as we do now, even though
the bird records themselves are getting vetted against the bioregional
filters. The big advantage here is that if we have a strong set of
bioregional filters, we catch most of the erroneous records more easily
(i.e., we can set filter numbers more specifically to those expected in
particular habitats not higher to accommodate a wider diversity of habitats
in a single county). And, by the way, the county level filters will still be
running in the background, but only operate if there happens to be a gap in
the custom bioregion filters, a rare event ideally. Creation of the
bioregional filters will require work an d we will need to coordinate their
maintenance because most of them cut across county lines (and therefore
current reviewer responsibilities).
Third, a hybrid approach is possible. And I think this option is probably
what we should consider carefully based on the content of some complaints
and worries I have been hearing from a few of the current Oregon eBird
reviewer team. In this case, we create the bioregional filters, but in some
particular situations where we have concerns that there is too much
variation in bird communities in bioregions that are large (think north vs
south coast, north vs south Willamette Valley, north vs south Cascades) we
CAN divide these at one or more points along county lines. The important
drawback, of course, is that each subdivision of a bioregion will need to
have its own filter. That's fine as long as we can get commitments from
folks to create and maintain all those filters. For filters that cut across
geopolitical review assignments, we will need to be sure we communicate with
everyone who has filter access so that we are not working at cross purposes
with occasional revisions.
Our decisions should be directed at what will help us catch as many data
mistakes as possible without creating the need for excessive numbers of
filters to develop and, very importantly, maintain.
So, to move forward, we need to:
1. Review the map I sent, taking a close look at your areas of current
review responsibility. If the lines are not in the places you would like to
see them in your county(ies), then fix them. Send me a photo of a hand-drawn
line on your county map if you like. We can work with something that simple.
The lines on the map we sent you represent the "official" ecogeographic"
regions (excepting, of course, the buffers along the coast and Columbia
River) to a large degree; these are typically determined by plant
communities but bird distributions do not always follow those same lines.
Let us know how to make it best for birds. Do take into consideration
locations of heavily birded areas and think about the best bioregional home
of those locations.
2. Tell me if you think the bird community in a bioregion crossing your
county(ies) is so different from the community in that bioregion of other
counties that you want to have your own bioregional filter within your
county (that means you are going to have to create it and maintain it, so
weigh that decision). Or tell me that you think the larger bioregional
filter cutting across your county could work ok for now.
3. There may be cases where unique areas (unusual habitats birded
frequently) are likely to trip filters so often that we need a special
filter for that site. Propose those for the group to consider, please.
Apologies in advance if this now sounds blunt (email is a sucky medium for
subtleties of human communication), but if you do not chime in, we will move
forward without you. Basically everyone at the Corvallis meeting (some 12 or
14 of us) agreed that adding bioregional filters would create a net gain in
data quality. Experiences of the folks in other states and countries who are
doing this all think it works better. So please inspect the Google Earth map
for your counties, consider how this could all work to improve data in your
review area, and let us know your thoughts or concerns.
I'm probably forgetting something important. I usually do.
Now, where did I leave my binoculars?......
Best,
Doug
Account management and archives: www.freelists.org/list/orebird
<//www.freelists.org/list/orebird> Reviewers
for Oregon Counties: http://tinyurl.com/n4k64ef ;<http://tinyurl.com/n4k64ef>
Account management and archives: www.freelists.org/list/orebird
<//www.freelists.org/list/orebird>
Reviewers for Oregon Counties: http://tinyurl.com/n4k64ef ;
<http://tinyurl.com/n4k64ef>