Utility and placement of multi-species comparisons?

We seem to be rehashing a previous topic on this observation of two mollusk species:
https://www.inaturalist.org/observations/2305520
So I want to ask for opinions:
Q1: Is there utility in uploading multi-species observations for the specified reason of offering direct comparison of two taxa? (That is, not just because an image inadvertently includes two or more species.)
Q2: If such observations are appropriate and useful, at what taxonomic level should they be placed?
Q3: Should such comparative observations be marked in the DQA as “No, the ID is as good as it can get.”?
My own take on these questions: Q1: I think carefully crafted/photographed comparisons are of great value in cases of similar looking species. Each of the subject species can and should be documented individually as well (i.e. to species level). Q2: The logical location for multi-species observations would be at the lowest common taxon, most often genus or family, for instance. Q3: While such multi-species comparisons are not appropriately moved to a lower taxon, the community can still support the placement at genus/subfamily/family level with appropriate IDs, so voting “No” to the improvement DQA unnecessarily kicks them to “Casual” status, IMHO an inappropriate result.

Opinions?

1 Like

I would duplicate the observation and ID each species to species level, noting “left individual” or something like this. You could also mention the other species in the notes. Then you have the photo with comparison of the two, but each species separate.

8 Likes

Well, I mean there’s an entire project for observations that show multiple similar species: https://www.inaturalist.org/projects/similar-species-in-the-same-photo-syntopy

It’s a bit different than the example in question, since the project is partly about the challenge of finding multiple species in the same genus in a single location without manipulating the scene, but I’ve found some of the side-by-side plant comparisions quite valuable, so I don’t think such observations are useless.

There is no rule that says an observation can only contain one organism, as long as the organism of interest is present in all the photos.

But clear communication about what the observation represents should be key. I think you do need to decide whether you want the ID for the observation to reflect the lowest common taxon, or whether you wish to choose one or the other organism to be the focus of the observation.

In the former case, marking “as good as it can be” seems like a reasonable way to reflect the fact that you have chosen not to ID the individual species in the photo. Of course, if the lowest common taxon is at family level or above, this will have the effect of making the observation casual, so you will have to decide whether you are OK with that.

In the latter case, to prevent confusion for people who might be viewing the observation or the photo without further context, I think it can help in such cases to label the image itself in such a way that it is clear there are two different species (and which one is which).

If the shells were moved more than a few meters from the location they were found in order to get them both in the same photo, I guess you would need to mark the DQA as “location inaccurate”, but I don’t see anything that requires such observations should be made casual per se.

1 Like

Susanne, that might be one way of doing it, but one unintended consequence of keeping two species in one image and moving such an image to species level relates to the machine learning for iNat’s Computer Vision. CV doesn’t know “left or right” in a photo, even with labeled specimens. So in the case you set up, CV would be including the same identical image in training sets for two different species and trying to integrate what it “sees” to its identification algorithm for each species. The size of actual training sets will probably swamp out any such confusion, but it probably adds fuzziness (uncertainty) to the identifiability of each included species. This is a primary reason I think such multi-species images should not be identified or placed lower than the lowest common taxon (e.g., genus, family).

I would probably crop it to show one individual in first photo, and show comparison in later photo with note in comments.

3 Likes

In my case, the locations are manually added to the observation (rather than using GPS) and are placed at the location where found, not where photographed. So the locations are accurate (within the accuracy range I specify). It’s an advantage (?) I have in using simpler cameras!

1 Like

@bbk-htx,
sorry, I clicked the wrong button, I tried to cite you, because I think it’s a great solution! I would still hate to have two perfectly identifiable species at genus or family level. And nobody interested in the difference of the two species would find an observation like this.

1 Like

does the computer vision training only occur with the first photo in a given observation? i didn’t think this was the case. (the vision may make suggestions based on only the first image in an observation though.)

but i don’t think it really matters that computer vision may get trained on an occasional photo that has more than one organism in it. there’s no reason to keep an observation at a high level just to prevent computer vision from training on it.

2 Likes

Sorry, what I was thinking of was the fact that in a composed picture (as opposed to an in situ picture), the two shells could have been found at different places or times. So if you want the observation to represent the lowest common taxon, it would be difficult to assign place and time data that accurately applies to both.

Or if they are to be ID’d separately, you would need to adjust this data for each observation rather than simply duplicating the observation.

If you found them together or near each other, this consideration does not apply, of course.

1 Like

Each observation is for a specific organism, so in observations with multiple species in the same photo, the observer will ideally indicate which specific organism they wish to the observation to be of. The observation can be duplicated for other organisms in the photo, but shouldn’t be ided to the common ancestor of all organisms present. If this were the case, almost every anole observation (what I mostly do) would be at “Life” because anoles are almost always observed on/near plants.

The CV training can take place on any picture in an observation, not just the first. However, this doesn’t mean that one shouldn’t make observations with multiple species in them. The CV training will use a lot of pictures, so, unless there are a lot of observations with the same two species present, the CV likely won’t cue on that. This aspect of CV training can definitely be seen though - for instance CV results for many plant pathogens include both the host and pathogen as suggested IDs.

3 Likes

Susanne, I’ll reiterate that each of the species in the dual shell observation I linked to were also documented individually on the same date/location, so that data was not set aside or lost to research. My focus in asking these questions is the usefulness of photographic comparisons of two species and where/if they belong on iNaturalist. Let me offer my own research interests as an example: For several moth genera that I study, I would like to be able to access–on iNaturalist at the genus or family taxon level–side by side images of very similar species in order to compare and contrast them, but in nature that opportunity rarely happens. These types of comparisons are not uncommon on BugGuide, for instance, but the realm of that site is not as inclusive as iNaturalist and its collection of records is managed very differently. See, for instance,
https://bugguide.net/node/view/1461339
On BG at least, I seek out such comparative “observations” at the genus or family level, rather than requiring me to search for the same information on each/every species page. So on iNaturalist, I would hope to be able to search out such observations at the genus/family level.
Maybe I’m expecting something to suit my research needs that iNaturalist isn’t structured to host, and that’s fine. I just offer up such comparative observations on the very occasional moments when they present themselves and seem like they might be of some utility to someone!

Same beach, same day, same collecting effort.

No doubt are they useful! I have once made something similar: https://www.inaturalist.org/observations/131207760. But I made two observations for the two species. Just thinking: you could make a project which such observations, where they might be easier to find.

This conversation occurred also on this Turkey wing/mossy ark observation (actually it started there first)
I’ve put the link below so you can see the rest of the dialogue.

https://www.inaturalist.org/observations/2305669

I’ve seen many observations like this with multiple species of sea shells but never before on purpose and with both species being the intended recipient of the ID. Most observations that have two similar species are by accident. I usually try to pick the less common one and dedicate the observation to it. If the observer responds then I ask them to crop the picture but often times they never do which is why I have formed the unfortunate habit of making an ID first before asking intentions.

I have once seen someone post multiple species on purpose just to show the difference, but the default picture and ID were only for one species.

Actually that was me and I’m not sure if I want to post the link cause I’m a bit embarrassed by the exuberance I had to name a new species.
https://www.inaturalist.org/observations/124304737n
I did get to name the species in the database though! It only had the scientific name.

1 Like

Other taxon specialists cover this with a journal post.

iNat gets confusing if you are deliberately identifying two species in one obs. The taxon pictures are also confusing for the merely human. I can’t see ‘this ID is for the one on the left’

1 Like

I’ve posted comparisons of two similar species of grass a few times. (Two species in one photo, with note or comment highlighting the differences.) In each case, the observation was duplicated and one copy was posted for each species. I think such comparisons are useful.

I am not worried about training errors for the CV because (1) the comparison photos are a small percentage of the photos for each species, (2) the species they’re labeled as is indeed in the photo, and (3) the much bigger problem for CV training is observations of the two species that are mutually misidentified in iNaturalist. The benefits of training people on how to identify them correctly far out-weigh any harm from putting two species in the same photo.

3 Likes

I’d say definitely make two separate observations, one for each organism, and indicate which organism the observation is for. iNat observations record encounters with individual organisms.

2 Likes

I know someone who makes observations like this regularly. I think such observations are great and I often favorite them. Usually the person makes two observations (although the images are the same) so that we ID one observation as one species and the other observation as the other species.

4 Likes

Probably better to make 2 observations and refer to each in comment section