Gymnosoma rotundatum is an Old World species and is unlikely to be present in North America. There are six native North American species, all with similar appearance to G. rotundatum. They are difficult to distinguish from each other as well. ID to species from images is generally not possible, though there are some exceptions to this.
There are a number of observations of Gymnosoma rotundatum from North America. Absent more details and evidence of this species in North America, I think they should be reclassified to genus Gymnosoma. I have commented on research grade observations already, but here is a list of all observations:
https://www.inaturalist.org/taxa/460533-Gymnosoma-rotundatum/browse_photos?order_by=created_at&quality_grade=any&place_id=97394
More information, including lists of documented North American species, is at:
https://bugguide.net/node/view/4330
https://www.uoguelph.ca/nadsfly/Tach/Nearctic/CatNAmer/Genera/Gymnosoma.html
Feel free to re-identify them with comments and links to those sources.
It’s pretty common for things like this to happen. Someone identifies one in the wrong area for one reason or another, then the computer vision picks it up as being nearby, and recommends it for other species in the same genus.
Lots of old examples here although some have been fixed: Computer vision clean-up (archive)
Thanks–there are maybe hundreds of them.
I feel the prevalence of this issue of over-specific ID could use some help from the CV system. Many taxa, typically genera, are known to be difficult to ID to species from images. Has anyone proposed that a query be inserted into the CV suggestion saying “this genus has X number of species, and ID is difficult. Here’s one that looks similar, but you should check to be sure this is a plausible ID”. It seems like this could be a flag placed on genera where this is an obvious issue–many ground beetles (Carabidae), for instance. Any thoughts from the more experienced observers or curators?
Do all Gymnosoma species require dissection for species-level ID?
If you are certain that none of them are G. rotundatum, please identify them at Gymnosoma and click the orange button “No, this is not G. rotundatum, but it’s in Genus Gymnosoma.”
Looks like there’s 130 observations from North America, which is a fair amount, but using the ID page you could get through those relatively quickly with a copy and pasted sentence explaining that G. rotundatum does not occur in North America. I know nothing about the genus, otherwise I would help go through them.
As for the CV, that is another topic entirely. But getting all those observations to genus level would help improve the CV and is definitely worth doing.
Yes, that’s what I did for the RG obs. I will work on the other ones. Certainly I hope some of the “fly people” will jump in as well, perhaps adding more information, such as appropriate species ID’s if possible.
Edit. I’m not sure about all species requiring dissection for species-level ID, but I believe that several do. My level of knowledge of species in this group is limited to looking at references giving the range. A few observations on BugGuide have been made at species level, but I do not understand how that was done.
I’m no expert but I see it described as a “Palearctic” species, as well as posts from “Insect Sciences Museum of California - ISMC” on FB mentioning its presence in Petaluma (Sonoma County, California, United States)…
So if you are certain it’s not that one, maybe try to interfere in wiki and FB as well ;-)
I’m a bit surprised, because I thought only species that have been officially confirmed in a country would appear in the VR list (at least that happens to me with some species that are easy to ID but are not yet officially in my country, but this might be for other reasons)
Agreed, the best solution at this point is to sit down with it for an hour and push them back to genus in the ID modal. After that, the geomodel will update next month and stop predicting this species in North America. That should largely fix the issue going forward.
While it sounds like a lot, with a copy and paste saying this species doesn’t occur here, I bet you can get it done in less than an hour. I’ve done clean up projects like this where the number of observations number in the thousands before…
I just Explored this species in North America and found 130 verifiable observations, none of which have reached RG. I’ll take a stab at moving some of them back to genus level with a cross reference to the present discussion.
So I took about a half hour to go through all the Gymnosoma rotundatum uploaded from “North America” and moved about half of them back to genus or tribe level. I’ll leave it to someone else to pick up the rest. BTW, iNat’s CV is offering “Gymnoclytia” on a number of these observations and I’m not qualified to distinguish these two genera so I put those at tribe level as well.
Part of the issue is that observers are picking not the genus level ID suggested by CV but the first available species level ID, irrespective of CV’s confidence in it. This goes back to one of my fundamental beefs with the interaction of CV with the iNat community. I strongly believe CVs suggestion list should be confined to only taxa (and taxon levels) that it is highly confident of, e.g. 90%+. No lesser choices should even be listed. Even then we will see erroneous suggestions (and user choices), but the proliferation of suboptimal suggestions would be lessened.
Thanks, that’s great. There were no RG observations because I already commented on them. Part of my hesitancy with proceeding to comment on all those observations was me just trying to be cautious, not being an expert on Diptera. I was worried there was a taxonomic change or introduction to NA I was not aware of. I’m glad this became a group effort and not just me, jumping in without sufficient caution, something I’ve done before.
Gymnoclytia is similar to Gymnosoma. I managed to find my downloaded copy of the Manual of Nearctic Diptera, volume 2. Available to download here. Key for those two genera is on p. 1259; couplet 275:
-
Apex of scutellum without bristles; apical pair so widely separated and short that they appear as lateral bristles and cannot cross at their apices. Abdomen shiny, orange red with black markings, or sometimes entirely black, with short hairs each arising from a raised base, giving abdomen a somewhat rugose appearance. Pedicel and first flagellomere in most species elongate, extending to level of vibrissa (Fig. 87)
→ Gymnosoma Meigen; 6 spp.; widespread -
Apex of scutellum with the usual crossed pair of apical bristles (as in Fig. 190). Abdomen usually golden tomentose (in male) or gray tomentose (in female); intertergal sutures distinct. Pedicel and first flagellomere short, not extending to level of vibrissa (Fig. 88)
→ Gymnoclytia Brauer & Bergenstamm 6 spp.; widespread
Addendum. I think this sort of thing should be called the “Sorcerer’s Apprentice” ID problem. I feel like the CV model should have some way of accounting for this issue when it makes suggestions, and perhaps it does. I suggest some sort of score could be given at each taxon level indicating how feasible it is to ID the daughter taxa based on habitus images. Coming up with such as score would seem to be difficult. Perhaps some knowledgeable curators will chime in.
When a low-experience user proposes a fine-level ID:
Thanks for contributing your ID!
iNaturalist works best when everyone learns and grows. It looks like you haven’t identified many observations in this group yet. Your suggestion will stay visible, but it will need confirmation from others before it counts toward the Community ID.
Tip: If you’re not sure, it’s perfectly okay to suggest a higher level, like genus or family. Other users can help refine it from there!
In the Help/FAQ:
Why didn’t my ID change the Community ID?
iNaturalist values everyone’s contributions, but for very specific identifications (like species or subspecies), we rely more on input from users who have experience in that group. Once you’ve made more verified IDs in this group, your species-level IDs will start to count fully. In the meantime, your input is still visible to others, and helps start the conversation!
Great idea. I think an additional issue is, perhaps that those confidence assessments of the CV are not reliable. In this case it looked like the confidence increased over time, as so many suggestions were made, but the initial confidence was ill-founded. Users agreeing with the poor ID suggestion reinforces the system’s confidence, AFAIK. I saw a similar problem with species ID in the ground beetle Pasimachus as well–a few erroneous species ID’s appeared to snowball into many. However I do not understand the workings of the CV in this context. Perhaps somebody more experienced will comment.
Interesting points. Yes, Gymnosoma rotundatum is a Palearctic (Old World) species. I have not seen mention of it being recorded from the New World on BugGuide or in the Manual of Nearctic Diptera. It is certainly possible there are records from California–it would be nice to see a link to a particular Facebook forum.
I am not aware of any linkage of iNaturalist CV suggestions to official lists for a given country. Perhaps someone can correct me. AFAIK the suggestions are just based on what has been identified internally on iNaturalist.
I’m not sure there even exist reliable lists of insect distribution by country for most insects, excepting some well-known pests.
Another issue is that the appearance of the Palearctic Gymnosoma rotundatum is quite similar to that of known Nearctic species in the genus. I think you have to examine a specimen with a microscope and appropriate keys to get an idea. It might still not be feasible unless one has expertise in that group and access to research collections for comparison.
Strongly disagree with this suggestion. It moves away from the current democratic approach to IDs and creates a hierarchy that devalues IDs by certain users merely because they don’t meet certain arbitrary criteria. It is also likely to create confusion (why do some IDs “count” and others don’t) and I don’t see it as effectively reducing the number of users who make IDs without sufficient knowledge. In a situation like the one outlined in the original post, you have people making IDs and confirming IDs based on wrong assumptions about what species are present in the area. There is no reason to think that this would be prevented by requiring a certain number of a user’s IDs to be verified before they count. When the community collectively is wrong or misinformed, bad IDs will get verified, while according to your proposed system the IDs of a lone expert might never get the confirmations that would allow them to count towards the community ID.
I think a better strategy would be to refine the way the CV suggests IDs for difficult taxa and to develop a strategy to educate users about how to thoughtfully assess the CV suggestions. I’ve had numerous discussions with users who assumed that an observation had to be a particular species because the CV only suggested one or because there were no observations of other species in that genus in the area – when in fact there were other species not included in the model or there were lots of observations that nobody had yet managed to refine past genus.