[Tweeters] Novel machine-learning method produces detailed population trend maps for 550 bird species

[Steve Hampton]

John Fitzpatrick of Cornell Lab (retired now?) spoke in broad strokes about
eBird's Trends methodology at a WOS talk a couple years ago (during the
pandemic shutdown). I think it's not recorded.

The biggest challenge, as you might imagine, is that eBird use increases
about 30%/yr, or at least did at some point, and they are calculating
trends from 2007 to 2021. Many didn't start using ebird until about 2012 or
much later. So, using raw numbers, every bird is probably increasing
dramatically. They had to correct for a massive increase in observer effort
over time.

I remember they used presence/absence data only, not counts, so they're
focusing on the odds of detecting a bird on a checklist ("frequency").
That's crude, but over time probably approximates "abundance."  And, most
importantly, they only used eBird checklists from certain experienced
observers going to the same places repeatedly -- and these were chosen by a
computer algorithm. The computer knows which species you should be seeing
with what frequency, and at some point it cuts off checklists that don't
meet some criteria. So basically they mined the data for the most
comprehensive checklists.

He implied their methods would be published, but I'm not sure if this is
that publication or a later one, and perhaps the original one is out there.
Let me know if you find it!





On Tue, Aug 15, 2023 at 1:15 PM Robert O'Brien <baro at pdx.edu> wrote:


> Wow, thanks.  I did not know that was so easily available.

> As a test I just compared Rufous vs. Anna's Hummingbirds

> No great surprise that Anna's are on the increase, most especially around

> urban areas. SeaTac and then PDX

> Sadly Rufous is the reverse-strong decrease everywhere.  I have certainly

> observed this here SE of Portland over the last 50 years.

> I could describe further, but very easy for interested birders to check

> for themselves

>

> But, do you know of a scientific paper or report that describes how they

> arrive at the trends?  Correction for observer effort, accuracy and

> precision over time, etc.?

> Thanks again, Bob OBrien  Portland

>

>

> On Mon, Aug 14, 2023 at 1:16 PM Steve Hampton <stevechampton at gmail.com>

> wrote:

>

>> Ah, this is the science paper behind eBird's amazing Trends maps. They

>> are remarkable. In some instances I was able to compare them with other

>> data (e.g. surveys or even CBC data) and they lined up pretty well.

>>

>> To see them, go to *eBird*, click the *Science *tab, then *Status &

>> Trends*, and then type in a species name.

>>

>> It defaults to the *Abundance *map, which is a basic range map color

>> coded for abundance -- by season. You can then click on *Trends *and see

>> the red, white, and blue dots. If you hover over them, the detailed trend

>> appears.

>>

>> For example, if I hover over the dot nearest Seattle for Spotted Towhee,

>> it says declined 28%. That is since 2007 (there's a legend on the upper

>> right).

>>

>> These are wonderful maps -- each dot on each map is like a survey of its

>> own.

>>

>>

>>

>> On Sun, Aug 13, 2023 at 2:01 PM Alan Roedell <alanroedell at gmail.com>

>> wrote:

>>

>>> Interesting article. My impression after living in Seattle for 80 years,

>>> is that most species have declined.  Crows and accipiters buck the trend.

>>> I miss the swallows most.

>>> Alan, Seattle

>>>

>>> On Sun, Aug 13, 2023, 12:52 PM Dan Reiff <dan.owl.reiff at gmail.com>

>>> wrote:

>>>

>>>>

>>>>

>>>> https://phys.org/news/2023-08-machine-learning-method-population-trend-bird.html

>>>>

>>>>

>>>> Sent from my iPhone

>>>> _______________________________________________

>>>> Tweeters mailing list

>>>> Tweeters at u.washington.edu

>>>> http://mailman11.u.washington.edu/mailman/listinfo/tweeters

>>>>

>>> _______________________________________________

>>> Tweeters mailing list

>>> Tweeters at u.washington.edu

>>> http://mailman11.u.washington.edu/mailman/listinfo/tweeters

>>>

>>

>>

>> --

>> ​Steve Hampton​

>> Port Townsend, WA  (qatáy)

>>

>>

>> _______________________________________________

>> Tweeters mailing list

>> Tweeters at u.washington.edu

>> http://mailman11.u.washington.edu/mailman/listinfo/tweeters

>>

>


-- 
​Steve Hampton​
Port Townsend, WA  (qatáy)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman11.u.washington.edu/pipermail/tweeters/attachments/20230815/c0fe05b0/attachment.html>