Lemna minor: an example of what happens when there is no taxon expert tending to identifications

If you look at the Duckweed observation for Texas, you’ll see this: https://www.inaturalist.org/observations?place_id=18&subview=map&taxon_id=54519&view=species.

The top species identified by far is Lemna minor. But the Atlas of Vascular Plants of Texas lists collections for only 3 counties, which would make it the rarest species in the state if those records are valid. An expert from Alabama apparently stated that the species doesn’t actually occur in Texas. There are 530 iNaturalist observations identified as this species, 211 of which are research grade! It seems highly likely that these are almost all entirely misidentified. I’ve seen several other plant taxa where this situation is present. When the experts are there to make the identifications, iNat is great, but when they aren’t it can result in a mess. Thoughts?


Wow, that’s a lot. I’ve only come across a few examples of people picking common names that don’t identify a particular species (which I assume is mostly what’s happening here), luckily. Fixing several hundred entries like that is going to take hours of corrections and comments to the incorrect identifiers… that’s rough :disappointed_relieved:

I doubt it would ever be implemented, but it would be nice if there was a way to flag IDs on commonly misidentified taxa (obviously with a system built in for certain people to not see the flag) so that maybe some sort of warning pop-up would briefly explain how certain someone has to be on species and force an identifier to click something that would kick-out IDs where the person doesn’t know the binomial name, or something. Not really sure how such a system would work precisely, but it could be nice to have.


It happens. And it even happens for cases where at one point it was correct. When I have extra time I go through eastern NA Stereum sp because many are S ostera and now it is confirmed to be an Asian taxon and it’ll be (basically) one of three other species here - and usually it is IDable from photos. But look it up - there are thousands in the wrong locale! And because the incorrect was accepted for the past so many decades, many IDers don’t know it yet either. A lot of taxa lately develop far more quickly than info can be disseminated, especially for those of us without consistent journal access. Add to that this is a citizen science site - folk who dont even have such access ever. Plus, ID books that are old and outdated as soon as printed these days…you can’t say it’s just an iNat issue really, at least, I don’t think so.

We all find out lil pet projects to help along - there will always be more observations than there are IDers - both here and in the academic/“professional” community. Especially as universities stop teaching taxonomy which is a huge issue at least in the US and I believe Europe as well, generally speaking. Everyone who does things in the field know this: any research looking at taxa needs to actually check the sources.

Nature of science, not just the nature of this site ;)


Lemna is one of those taxa that also has the problem of being hard to ID from photos. All the keys I’ve seen require stuff like looking at the underside of a plant or counting veins in the leaf. iNatutalist users want to attach species names to everything, but for a lot of plants, that’s a bit of a gamble if you can’t take a closer look or don’t have material that’s reproducing.


This is a big reason why I’m extremely grateful to have some good taxonomic experts working in Bees. The value of the platform would be considerably reduced without experts.


This forum post and the older version of it have lists of many species where this is an issue: https://forum.inaturalist.org/t/computer-vision-clean-up-wiki-2-0/40318


It would be nice if curators had a tool that we could use to manually flag problem-species like this in such a way that the CV would be forced to suggest a higher-level taxon (genus, I guess would be appropriate here, though maybe “Lemnoideae” would be better) instead of aggressively prioritizing species-level misidentifications.


I wonder if there is a feature request for disabling certain taxa in CV? Region-blocking certain CV suggestions would also be helpful, but would definitely need to be applied sparingly and only by staff.


This would be fantastic:


I remember when I was first starting out on inat that I would always just pick the top and relatively similar looking recommended taxa. It would result in a budding sea thrift being marked as “genus Douglasia” (that now has an inactive taxon) which doesn’t even grow here in the UK at all! Being someone who was surrounded in sea thrift all the time, I am sure I put down a good 6-7 of these observations and marked them wrong. I would not be surprised if that was what was happening to this taxon as well.

1 Like

But are those RG observations? There’re many Lemna iders, so I don’t know this problem, they’re reided to genus pretty fast.

Try convincing your expert friend to become a iNat user!


I looked at the observations you linked in your post. It seems obvious, the smaller organism, the more detailed (zoomed) picture is necessary for proper identification. As a moss-observer I know the problem well. For proper moss ID sometimes you need microscopic slides and leaf cells’ shape.
It seems a lot of observations of your Lemna minor are just photos taken from a distance. iNat algorithm tries to suggest and force the ID on species level. And every new mistake confirmed by second user makes things even worse, as the database suggests the wrong species for next observations nearby. I can’t imagine how to solve the problem without good human-identifiers, who can deny species level of ID and go into just genus-level (or higher) ID.
Interesting thing about bryophytes - we have good and engaged identifiers in Europe (who try to make IDs constantly, every day), but if you jump just outside European border (for instance: Israel) there are many obvious mistakes in mosses ID - I mean even such mistakes like lichen identified and confirmed as “moss” by several local users.


And that’s the US where you have the best representation of experts on iNat. And vascular plants, one of the best-studied groups of living things. In other parts of the world the situation is much worse, especially with not so well-known groups. The true is that iNat is not doing enough to attract more experts and to make them interested to solve such messy stuff, instead their interests often take some third or fifth place. And of course it’s also a result of overestimating AI, which is only a tool that should be used properly, not a replacement of an expert, which is not explanatory on iNat, so of course people often using AI-suggestions without even checking if it’s actually similar.


It’s not just a problem of inconspicuous species either. Solidago canadensis is a popular computer-vision suggestion throughout western North America, where it does not occur except maybe as an ornamental escapee. There are 585 observations of in my home province of British Columbia, where it is a non-native taxon known from a single specimen. Goldenrods all look the same to many observers and to computer vision, and they reinforce each other. It doesn’t help that the name used to be a catch-all for what are now several species.


If you know what you’re doing and need to bump something up to a higher taxon by way of people putting IDs of “no, the evidence is not sufficient to conclude it’s [more specific taxon]”, I wonder if it may be useful to get a brigade of users to help do this.

But such an approach does feel a bit…vigilante and somewhat questionable.


Lemna, grasses, many others – identification depends on little structure that are rarely photographed clearly. Very often, all I can say is, “Probably not Lemna minor, but I can’t be sure.” That’s not satisfying for anyone and after all it (or some of them) might be Lemna minor for all I can really tell. So I don’t comment. Very frustrating.


I know folks will disagree, but in my opinion, the value of iNat are the observations themselves. If a duckweed researcher is interested in these observations, then they can hop on board and curate them. Until then, they’ll be here waiting for that moment to arrive! It’s OK if there are messes…keep the observations coming!

If a researcher comes along and thinks iNaturalist (or GBIF) is a repository of perfectly identified specimens and they don’t have to put any work into it, then they need to open their eyes to the more-than-obvious reality.


Taraxacum is full of ‘common dandelion’ but in reality is an insane genus and very difficult to ID. I wonder how many are misidntified but I woudn’t know the first thing about it.


If you know what you’re doing, then this is exactly how iNaturalist works - a process of continual refinement and correction. I agree that it might be somewhat questionable if those who don’t know what they’re doing start going in and correcting these IDs without adequate knowledge, although it’s surely no worse than those who have already added overly fine IDs without adequate knowledge…