How Good are we at Counting? - 22 Feb 2010
"I've been playing with Photoshop some more and came up with another super-panoramic image of the Detroit River at the foot of Southfield Rd. in Wyandotte (Dingle Park). On the 30th of January there was a large raft of Canvasback/Redhead Ducks that I photographed using the D300 and 400mm Sigma. I stitched together approximately 30 images to produce this image: I performed some rough cloning on the top and bottom of the image to fill in dead space so if it looks a bit unnatural that's ok.
http://www.pbase.com/jourdaj/image/121761235/original
Now comes the challenge: I'm curious about what kind of numbers you folks would come up w/ trying to guesstimate the numbers of ducks/geese/swans in the image. Use whatever mental filters you have to get an approximate number in 60 seconds or so. Please respond off-line if you choose to participate - I'll summarize results (leaving out names) and report back to the list.
I tried this and came up w/ a number, which I'll reveal later. I then took the original, TIFF image I created (805 MB in size), blew it up, and hand-counted every single bird in the image (don't ask me why....). I was curious to see how good our perception is at trying to determine large numbers of birds when we gather census data. How big is our systematic error? I managed to get 90% accuracy when I tried it."
So here are the results:
64 Canada Geese
289 Swans - estimate 95% Mute (275) and 5% Tundra (14)
10 Great Blue Heron
9 Gadwall
7 American Wigeon
11 Goldeneye
5 Bufflehead
7 Hooded Mergansers
12 Ring-necked Duck
16 Lesser Scaup
4 Common Merganser
7 Ruddy Duck
6 Mallard
534 Redhead (originally estimated ~10% of mixed Redhead/Canvasback)
4058 Canvasback
---------------------------
5034 Total
I originally Estimated 4500 total birds based on 60-sec. count for 89% accuracy. Others who played along had up to 95% accuracy based on 60-sec. count.
Those that participated (7 total) submitted the following range of counts:
833 (low) to 7200 (high)
Two participants were 95% accurate with counts of ~5300 (Joe Hildreth and Cendra)
Three others were 75-85% accurate w/ counts of 3750 - 4250
Surprisingly, a 60-sec count of large numbers like this are quite accurate. I would expect generally lower numbers than 5000 since so many of the ducks were obscured and so closely-packed. Though it may not be the best example to test your skills it is an image that we encounter from time to time and have to do the best we can with.
Thanks to all who played along!
Ecorse Park, Ecorse US-MI (42.2378,-83.1466), Wayne, Michigan, US
Jan 30, 2010 3:00 PM - 4:00 PM
Protocol: Stationary
Checklist Comments: I've been playing with Photoshop some more and came up with another super-panoramic image of the Detroit River at the foot of Southfield Rd. in Wyandotte (Dingle Park). On the 30th of January there was a large raft of Canvasback/Redhead Ducks that I photographed using the D300 and 400mm Sigma. I stitched together approximately 30 images to produce this image: I performed some rough cloning on the top and bottom of the image to fill in dead space so if it looks a bit unnatural that's ok.
http://www.pbase.com/jourdaj/i
Now comes the challenge: I'm curious about what kind of numbers you folks would come up w/ trying to guesstimate the numbers of ducks/geese/swans in the image. Use whatever mental filters you have to get an approximate number in 60 seconds or so. Please respond off-line if you choose to participate - I'll summarize results (leaving out names) and report back to the list.
I tried this and came up w/ a number, which I'll reveal later. I then took the original, TIFF image I created (805 MB in size), blew it up, and hand-counted every single bird in the image (don't ask me why....). I was curious to see how good our perception is at trying to determine large numbers of birds when we gather census data. How big is our systematic error? I managed to get 90% accuracy when I tried it."
I originally Estimated 4500 total birds based on 60-sec. count for 89% accuracy. Others who played along had up to 95% accuracy based on 60-sec. count.
Those that participated (7 total) submitted the following range of counts:
833 (low) to 7200 (high)
Two participants were 95% accurate with counts of ~5300 (Joe Hildreth and Cendra)
Three others were 75-85% accurate w/ counts of 3750 - 4250
Surprisingly, a 60-sec count of large numbers like this are quite accurate. I would expect generally lower numbers than 5000 since so many of the ducks were obscured and so closely-packed. Though it may not be the best example to test your skills it is an image that we encounter from time to time and have to do the best we can with.
16 species
Canada Goose (Branta canadensis) 64
Mute Swan (Cygnus olor) 275
Tundra Swan (Cygnus columbianus) 14
Gadwall (Mareca strepera) 9
American Wigeon (Mareca americana) 7
Mallard (Anas platyrhynchos) 6
Canvasback (Aythya valisineria) 4058
Redhead (Aythya americana) 534
Ring-necked Duck (Aythya collaris) 12
Lesser Scaup (Aythya affinis) 16
Bufflehead (Bucephala albeola) 5
Common Goldeneye (Bucephala clangula) 11
Hooded Merganser (Lophodytes cucullatus) 7
Common Merganser (Mergus merganser) 4
Ruddy Duck (Oxyura jamaicensis) 7
Great Blue Heron (Ardea herodias) 10
View this checklist online at https://ebird.org/checklist/S4
This report was generated automatically by eBird v3 (https://ebird.org/home)
Jan 30, 2010 3:00 PM - 4:00 PM
Protocol: Stationary
Checklist Comments: I've been playing with Photoshop some more and came up with another super-panoramic image of the Detroit River at the foot of Southfield Rd. in Wyandotte (Dingle Park). On the 30th of January there was a large raft of Canvasback/Redhead Ducks that I photographed using the D300 and 400mm Sigma. I stitched together approximately 30 images to produce this image: I performed some rough cloning on the top and bottom of the image to fill in dead space so if it looks a bit unnatural that's ok.
http://www.pbase.com/jourdaj/i
Now comes the challenge: I'm curious about what kind of numbers you folks would come up w/ trying to guesstimate the numbers of ducks/geese/swans in the image. Use whatever mental filters you have to get an approximate number in 60 seconds or so. Please respond off-line if you choose to participate - I'll summarize results (leaving out names) and report back to the list.
I tried this and came up w/ a number, which I'll reveal later. I then took the original, TIFF image I created (805 MB in size), blew it up, and hand-counted every single bird in the image (don't ask me why....). I was curious to see how good our perception is at trying to determine large numbers of birds when we gather census data. How big is our systematic error? I managed to get 90% accuracy when I tried it."
I originally Estimated 4500 total birds based on 60-sec. count for 89% accuracy. Others who played along had up to 95% accuracy based on 60-sec. count.
Those that participated (7 total) submitted the following range of counts:
833 (low) to 7200 (high)
Two participants were 95% accurate with counts of ~5300 (Joe Hildreth and Cendra)
Three others were 75-85% accurate w/ counts of 3750 - 4250
Surprisingly, a 60-sec count of large numbers like this are quite accurate. I would expect generally lower numbers than 5000 since so many of the ducks were obscured and so closely-packed. Though it may not be the best example to test your skills it is an image that we encounter from time to time and have to do the best we can with.
16 species
Canada Goose (Branta canadensis) 64
Mute Swan (Cygnus olor) 275
Tundra Swan (Cygnus columbianus) 14
Gadwall (Mareca strepera) 9
American Wigeon (Mareca americana) 7
Mallard (Anas platyrhynchos) 6
Canvasback (Aythya valisineria) 4058
Redhead (Aythya americana) 534
Ring-necked Duck (Aythya collaris) 12
Lesser Scaup (Aythya affinis) 16
Bufflehead (Bucephala albeola) 5
Common Goldeneye (Bucephala clangula) 11
Hooded Merganser (Lophodytes cucullatus) 7
Common Merganser (Mergus merganser) 4
Ruddy Duck (Oxyura jamaicensis) 7
Great Blue Heron (Ardea herodias) 10
View this checklist online at https://ebird.org/checklist/S4
This report was generated automatically by eBird v3 (https://ebird.org/home)
