Hello Chuck, Your comment that the photo of the Chestnut-collared Longspur (Calcarius ornatus) was photoshopped to remove vegetation obscuring the view of the bird piqued my interest. It seems to me that in some circumstances there are surely issues with photoshopping photos where the science of bird identification would be compromised. This instance may or may not be an example of a compromised identification. You cite three factors supporting the ID: 1) the experience and authority of the observers, 2) field marks observed when the bird flew, and 3) the photograph itself. Of course, ultimately acceptance of the sighting will be the responsibility of the appropriate reviewing committees. I would think they should require both photos of the bird, before photoshopping and after, to determine if photoshopping altered the field marks in any way that might call in doubt the crucial field characters of the bird.

I’ve been using a recent version of Photoshop with generative AI capabilities that very quickly can delete distracting imagery from the main subject of the photo. I have done this for a variety of photos — basically experimenting — including photos with birds, animals and people. So far, I have achieved some marvelous results on removing background distractions such as a large branch in the background behind the subject, or removing the image of person from a photo that was inadvertently included in the view because of a framing failure in the first place. Where Photoshop’s AI tool had problems it involved removing unwanted images of persons cluttering up the background, but where the background was “generic vegetation” the results were spot-on. Of course, AI in working with images relies on a huge “library” of stored images which it uses to predict what should be an optimal replacement as well as using elements of the photo it is working on to achieve an acceptable and realistic result. In the case of a rare bird, where AI’s bank of imagery might be limited, it could easily obscure a critical field character or not find valid match. AI programs in the case of text, either editing it or generating it, has been known to “hallucinate”. Surely this is true with imagery!

I’ve also worked with older versions of Photoshop where a lot of “manual” manipulation of its tools is required, e.g., painting out an intrusive branch crossing through the beak of a bird using pixels copied from a visible portion of the beak. I have some success using these tools but it is very time consuming compared to what generative AI can do in seconds. And I would challenge anyone to say that the photo was manipulated in any way, same with those altered by generative AI.

Seems to me that this raises some ethical issues especially in cases where a photo is the determining variable of a bird’s ID, for instance in the case of discriminating subspecies, not just the rare or seldom seen bird. It is incumbent on photographers, or will be, to state to what extent a photo has been altered.

Photoshopping in general involves all sorts of manipulations, from improving exposure and contrast to “enhancing” colors and hues. We’ve all seen photos of the same bird taken by the same camera, and by different cameras, and have noted a wide range of differences between the resulting photos. That probably is why in some cases rare bird committees don’t just rely on photos but require written notes.

I am gratified that you stated that the photo of Chestnut-collared Longspur (Calcarius ornatus) was manipulated by a photoshop program. I am thinking that as a matter of common practice, say in exhibiting our photos we state to what extent we have altered the image. In the case of “citizen science” projects it is imperative that we do so.

Thank you!

Thomas J Hinnebusch

LikeLike