Monday, August 21, 2006

Luneau: Basics and the The Underwing

Jumping ahead a bit here, but there are some fundamentals about this that are rolling loudly in my head...

First, some starting points for understanding the black and white of the video:

White bleed, focus, etc.: Much reference has been made to the video exhibiting "white bleed," a contamination by excess white of the pixels surrounding a bright white pixel. But look at the edges of the tree trunks as they pass in front of bright sky, then gray background trees, and you will see that there is not really any white bleed. It has also been said that the video is poorly focused. Close up examination of the patterns of tree trunks shows that this is not the case, either. What the video is, pure and simple, is pixelated, nothing more. The illusion of poor focus is an artifact of the video presentation, in which the pixel edges are softened to make the image look less jarring and jagged. The pixels on that bird are HUGE, about 3 cm across early on, and larger later. This simple pixelation is plenty adequate to account for all the crappy resolution of the images. The other factor killing resolution is plain-and-simple motion blur. This is what causes the white expanse on the underwing to appear larger than it actually is, and obliterates details of the underwing. Go through the arithmetic:

The pixel resolution is about 3 cm or worse. To this add motion blur of several more centimeters on the wings (more than 10cm near the tips in some frames). We're not going to see things narrower than 5-10 cm that are aligned perpendicular to the wing's motion. Now look at the sizes of the underwing features that would differentiate the two species: The Pileated black trailing edge is about 3-5 cm wide; the Ivorybill medial black bar is of similar width. And these wings are not oriented face-on to the camera, so the apparent width of these features is even less than this. So of course these features are completely unresolvable in this video. Either species would be expected to show a big white blur under its wing and nothing else. In the CLO analysis they argue that the Ivorybill black bar would be obliterated, but the Pileated black trailing edge should be visible. Given that the two features are of similar width, this is argument does not hold up.

The one black feature of the underwing that is visible consistently is a black wingtip. This is especially clear in the second half of the downstroke of many frames early in the flight sequence, where it is smeared into a consistent black arc. Bevier and others have argued that this is a specific Pileated characteristic, and its presence is a key point they use to conclude that the bird is a Pileated. This makes no sense, however. Both species have black wingtips when viewed from below. The distance from the wingtip to the distal end of the underwing white is quite similar in both species, comprising about 20-25% of the total wing length. I do not understand, therefore, why the presence of a visible black wingtip provides any information at all to identify this bird.

Conclusion (and starting point for further analyses): The underwing pattern of the flapping wings in the Luneau video does not provide any information at all to help determine the species of the bird. Both species would be expected to show white underwings with black wingtips and no other markings under these conditions.

But what about the black wing edges that are visible in these early frames? Other than the wingtips, they are artifacts. They are not consistent between frames, and they appear both on the trailing and leading edges of the wings with approximately equal frequency (neither species should show black on the leading edge of the underwing except at the tip, where both should show it). They also frequently align with darker areas in the background. The blurred wing images are semi-transparent, and they often meld with the background deceptively. The black spots on the wing edges are garbage, noise, useless; not information.

So, I have just thrown out the most prominent plumage characteristic in the video, the big fat white underwing with a black tip, concluding it is useless for identifying the bird. What is left? Good question. I don't think (yet) that the answer to that is "nothing;" but I make no guarantees.


At 2:13 PM, Anonymous Anonymous said...

I'd make some comments concerning the black wing tips.

Both species have black wingtips when viewed from below.

Sort of true. IBWO should show a partial black wing tip, due to it having 5 mostly black primaries from below. PIWO shows a curved black wing tip since all primaries have dark tips.

Look again at how the black curves around the entire tip of the wing in the video. For this to occur, the black would have to occur on all or nearly all primary tips. I can't see how this pattern of black could be created by 5 dark primaries, while the white primary tips just disappear.

The distance from the wingtip to the distal end of the underwing white is quite similar in both species, comprising about 20-25% of the total wing length.

I don't believe that this is not as important as the number of primaries with black tips, but I do have a comment on it. You mention total wing length. What I'd like to know is percent from tip to wrist. If IBWO simply has proportionally longer inner wings, your comparison doesn't hold up. In the images I can find, it seems that the black on IBWO comes closer to the wrist than PIWO, making the pattern in the video worse for IBWO.

Here's a hypothetical example. If PIWO has a proportionally shorter inner wing and primaries that are half black, and IBWO has a proportionally longer inner wing and primaries that are 3/4 black, the black wing tip of IBWO should be much more extensive appearing.

If you have access to specimens or good underwing shots, I'd like to know if this is a consistent difference between the two species.

I do not understand, therefore, why the presence of a visible black wingtip provides any information at all to identify this bird.

Again, back to the shape / extent argument. I don't think that these two species should show the same black pattern on the underwing.

Finally, look at the David Nolin video of multiple PIWO on In takeoff flight, the black trailing edge is actually pretty hard to see even with good video. The curved dark wingtip, however, is quite clear and resemble the Luneau video to me.

Great post, BTW.

At 2:42 PM, Blogger Bill Pulliam said...

If you look at all the frames in sequence after the bird emerges from behind the tree, you will see black pixels appearing all around the wing: leading, trailing, and tip. I don't feel it is possible to infer the shape of the black blob at the tip of the wing from this; I am only comfortable inferring that there is a black area at the wingtip. There is just too much spurious black for me to be confident in anything stronger than that.

You mention total wing length. What I'd like to know is percent from tip to wrist.

Most illustrations of the species show the white along the leading edge of the underwing extending farther past the wrist in the Pileated than the Ivorybill, but only by a matter of a few centimeters. The white on the trailing edge is generaly shown to extend significantly farther distally in the Ivorybill than the outermost extent of white in the Pileated underwing, with the result that the "black blob" on the IBWO underwing is situated closer to the leading edge than on the PIWO. These differences are unlikely to be distinguishable with 3-cm pixels on a rapidly moving wing, however. Unfortunately I have no museum access, so I will have to defer to those who do to make these determinations.

One of the goals of my animations is to take the sketches of wings frame by frame and attempt to fill in the black/white patterns of both species. I think this will give a clearer idea of just what out to be visible after blurring and pixelation. The thing about it is, these two species really are not all that terribly similar. The underwing patterns are strikingly different. The difficulties in extracting a definitive idenification from the Luneau video speaks very strongly as to the extremely poor quality of this video!

I would strongly prefer, by the way, if you signed future comments. I really dislike anonymous postings in serious discussions.


At 7:27 PM, Anonymous Anonymous said...

Bill, I agree that "bleeding" is not happening due to light spilling over to neighboring pixels or other purely electronic artifacts. However because of motion blur from the slow shutter speed and low res each CCD element is capturing a sample of the photons reflected off the moving wing. If black and white pass through the field of view of an element during the 16.7 ms of exposure the white will likely overwhelm the black to produce a lighter shade of gray. In this way white is more dominant in the stills and will grow in size in comparison to the size white in an image at a higher shutter speed. Black that neighbors white along the path of motion will tend to shrink. In this way white will "bleed".

Bill, I agree with attempting to determine wing tip shape from the video given the resolution and the motion blur is dubious. However I think that the extent of black at the wingtips is quite different between the 2 species. If you look at the photographic evidence of IBWO in flight there is actually not a lot of black at the tips and it is likely neighbored by white primaries along the path of motion (see the implications of this above in "bleeding").

Take a look at the available flight photos on Cornell's site:

In particular look at the slightly motion blurred images at the bottom of the page. These images do not show a prominent black wing tip, unlike photographs of PIWO wings.

Look at Eckleberry's IBWO sketch in "The Grail Bird." The white of the primaries comes out very close to the tip of the wing. Put that wing in motion at low res and a slow shutter speed and I don't think you would see an extensive black wing tip because the white primaries will overwhelm the black primaries.

I've heard that the amount of white in the primaries on the skins in collections is variable. It would be interesting to see how this actually varies to better understand the implications relative to the visibility of black wing tips.

At 7:58 PM, Blogger Bill Pulliam said...

re: bleeding/motion blur. I agree completely, and this should be sufficient to obliterate the underwinf pattern of either. If anything, the black PIWO trailing edge would be more likely to be resolved, as it will only be contaminated by the gry background rather than the white underwing. However, I still doubt a narrow feature like that would register in htese frames either.

re: wingtips. The black wing tip, where it registers as the black arc at the tip of the moving wing, is only about one pixel wide, or about 10% of the total wing length. Again, I think it is highly iffy to try to resolve any species-level difference from a blur of one pixel width. I think we must be consistent in interpretation: if we accept that the resolution is too poor to consistently show a PIWO black trailing edge, we should not then attempt to infer differences of only a few cm in the size or shape of other features.

The extent of white on the IBWO underwing in published illustrations varies quite a bit between them. The two useful Tanner photographs (which could well both be of the same individual) appear show a bird with even more white than in these illustrations. I am assuming that the artists used study skins as references for their work; hence it appears that the individual IBWOs probably do vary in the amount of white.


Post a Comment

<< Home

Site Meter