Realistic comparisons



[This page is under construction. I intend to discuss several issues along the theme below. Many of these question remain confused by those expressing opinions in the literature and elsewhere]


What are reasonable approximations of reality?


Are the depictions by Fitzpatrick et al. (sketches [2005] and photo montages [2006]) reasonable approximations of reality for an hypothesized Ivory-billed Woodpecker perched more or less upright on a tree trunk?


  1. • Use of digital still images of Pileated wings cropped and pasted into photographs versus images derived from interlaced, out-of-focus video at a distance


  1. • Use of digital still images of a cropped Ivory-bill wing taken from an extreme ventro-caudal angle unlike the position hypothesized; this cropped image itself modified by omitting most of the black forewing undermines the measurement of “wrist to tail tip” crucial to the proponents arguments


  1. • Comparisons to Pileated video and wings broadside to the camera rather than flying away and slightly above camera level


  1. • Reasonable approximations as interpretive sketches: does adding a black bar on the underwing, white basally in primaries, and details of white tips to inner primaries accurately depict what is present in video (see Fitzpatrick et al. 2005, figure 2A and 2C)


  1. • The issue of wing twist (and camber). This has been documented in several papers by Bilo (1971, 1971, 1981). The House Sparrows that he studied have a wingbeat style that is quite similar to woodpeckers. Hedrick et al. demonstrate how angle of attack and rotational circulation (which implies long-axis rotation of the wing) varies during the wingbeat cycle in cockatiels and doves.


  1. • white blur (focus, motion, interlaced vs. deinterlaced); artifacts versus real patterns


  1. • Accurate descriptions of take-off mechanics and timing: e.g., pertaining to the elevation of the wings when a bird is alarmed or seeking to rapidly take off and gain altitude. In general, I suspect that woodpeckers and other flap-bounding birds will keep their wings closed if they are translating potential energy for kinetic energy at the start of a take-off.  A bird that is startled, or one that is seeking to immediately climb (like the one in the Luneau video), will tend to lift its wings immediately. Papers to support this prediction include one by Earls and another by Tobalske that demonstrates that motivational state has an effect upon the timing of wing elevation. In the Luneau video, we have woodpecker perched low over water and needing to climb and escape.  Hummingbirds when startled will lift their wings and get them beating earlier than when they are taking off in a calm manner or when they are taking off to chase a conspecific.



Anatomy of a video frame


It is important to understand how digital video captures (acquires) a scene in interlaced mode. The first image below shows a 10x view of the raw digital information that comprises the video fields labeled 16.7 and 33.3 by Fitzpatrick et al. An entire video frame is 1/30th of a second, a relatively long period of time for any moving object to be captured accurately. During this time, the camera records first one set of alternating lines (odd) in 1/60th of a second and then the other alternate set of lines (even) in the last 1/60th of a second.


To analyze the image, Fitzpatrick and others deinterlaced the digital image. These are represented in the second dual set of images, here modified to show the alternating lines actually captured. An important point to note is that every other line contains no data for each 1/60th of a second scan.


When the video frame is deinterlaced, an algorithm is applied to estimate the missing lines and fill in the field to make a complete image. The algorithm commonly chosen, and used for the Luneau analysis, is interpolation.


At this point of the deinterlacing-zoom reconstruction, 50% of the data being debating is an estimate. It gets worse than that because to zoom the image another algorithm is used to multiply the existing pixels and estimate what color they should be. To make the 10x image of the bird in the Luneau video, a bicubic resampling method was applied. This means that for each pixel, visible as individual squares in the first image below, 10 more are created to “smooth” the zoomed image. The final result is in the third image below.



“Raw” data, actual pixels, showing 10x zoom of a single frame from the Luneau video. The data here comprise what become the smoothed images labeled video fields 16.7 and 33.3 (numbers that represent time in milliseconds from the beginning of the segement of the Luneau video presented by Fitzpatrick et al. 2005).






video field “16.7” on the left and “33.3” on the right. Each is composed of 50% computer estimates for every other scan line and ten-fold resampling estimate to smooth pixels.



Launch sequence


Video of Pileated Woodpecker captured with same settings and focus as Luneau bird confirm patterns and motion seen are identical.






Video of a Pileated Woodpecker (upper two rows) as seen at one-third the distance (~6 meters) of the Luneau bird (bottom row; 20 m) shows that timing and mechanics of wing movements match. Note position of tail (orange arrows) and near lack of black trailing edge (blue arrows), even on a bird much closer than the Luneau bird. Note also that full wing extension is visible for two full video frames (each 1/60th of a second); this matches positions of seen for Luneau bird. Also note how initial movement of wing in preparation for launch is dorso-medial movement of humerus, revealing white underwing immediately. Primaries begin to spread outward after this lifting motion of wing at wrist. This contradicts assumption that primaries spread and underwing remains hidden as suggested by Fitzpatrick et al. 2005.