Making Meaningful Identifications

There are lots of users going through and just identifying plants as "Plants" (Plantae and Tracheophyta). That is hardly useful, and does not really contribute anything of value to the identification process. It would be better to leave them as unidentified until someone knowledgeable put then in a useful bin - ideally Family or lower.

However that does not mean that they cannot be meaningfully identified.

There are three major areas where you can assist in moving these token identifications to a more useful identification.


Trees are easy to identify as trees, and then a lot harder to ID further.
However, the ID of "Tree" is extremely useful as fewer than 10% of our plant species in Africa are trees, and there are specialists, enthusiasts and locals who know their trees better than any other plants. And their are field guides to trees, which assists in further identification.

The easiest way to facilitate this, is to add a tree observation to a "Tree Project"
We do have one for southern Africa, which despite its name, has no geographical restriction. Although based on our field guides, it would perhaps be best to keep it to southern Africa, and create a new one for our other floral regions. On the other hand, it is so easy to filter by region, should we rather not just use this single project or all our African trees?
Let us try the latter route: if it gets unwieldy, we can create more tree projects (or if you would prefer to do so anyway, please do so and tell us).

The project to add is
you need to first join it, and then you can add it to any observation (except those users who restrict adding projects to their observations)

So instead of just putting a meaningless ID of "Plant" onto an observation of what is clearly a tree, please add a tree project (you can add the higher ID too if you want, but it is purely superfluous as the tree enthusiasts will almost certainly use the project first).
Here are some trees, that have been identified as trees, but not linked to any project.
(Unfortunately, the curation tool is not project friendly, and you will need to pop out the observation to add a project. (a way around this is to add a comment and later filter for comments and add the project).
These are identified as trees by the AI. Not all are trees, but most are. They have not yet been put into the Trees project.

Note that there is a tree project for southern Africa, based on identified trees. You may find it useful:
Details are here:


Please remember to identify Lichens as Lecanoromycetes (= Lichens) rather than simply is Fungi.
Yes, not all Lichens are in this group, but almost all of them are: so the specialists will reidentify in those few cases.

Observations with more detailed suggestions that can be easily checked

The AI has provided identifications for these observations to a level below their current ID on iNaturalist. Not all these IDs are correct, but most should be able to be taken to a further level - to where specialists can access them:
(only groups with more than 20 observations are listed)
[a million thanks to @jeanphilippeb - a pity iNaturalist cannot automate this identification to some degree - all the below are thanks to Jean-Philippe]

DICOTS 8,600 observations:

Mimosoideae 700 observations:

Euphorbiaceae 420 observations:
Fabaceae 410 observations:

MONOCOTS 390 observations:
Lamiales 380 observations
Poaceae 360 observations:
Polypodiales some fern knowledge required - 330 observations:

Crassulaceae 280 observations:
Asparagoidea 260 observations:
Combretaceae 200 observations:
Sapindales 200 observations:

Agavoideae 160 observations
Anthospermae 100 observations:
Hibisceae 100 observations:
Tetragonia100 observations:

Rutaceae 80 observations:
Campanuloideae 75 observations:
Vitaceae 66 observations:

Iridaceae 46 observations:
Eucalypteae 45 observations:
Mimoseae 42 observations:
Alismatales 40 observations:

Rosaceae 35 observations:
Pedeliaceae 34 observations
Urticaceae 32 observations:
Myricaceae 30 observations:

Scrophulariaceae 24 observations:
Malvales 22 observations:
Bombacoideae 20 observations:
Loganaceae 19 observations:

Posted on August 07, 2023 08:17 AM by tonyrebelo tonyrebelo


Hi, just so I understand correctly – we can add trees from all over Africa to the tree project, in spite of the name?

Posted by jensu 7 months ago

Yes - you are welcome.
I will add relevant filters.

Posted by tonyrebelo 7 months ago

Tony, immense thanks for organizing all the important topics in one cogent set at this stage in the challenge! I really was hoping to present the Unknown projects links in a journal post, and now I don't have to. :)

My additional advice to identifiers moving past Plantae(p) and Tracheophyta(t) is to consider getting to Angiospermae(a), even if you can't get to Dicots(d) or Mononcots(m). Angiosperms is where important flowering(pl)/fruiting(pr) annotation can start (p and t don't have those options). @astra_the_dragon related to above keystrokes and the wiki tutorial in the forum, I was planning to build that out from experience in this challenge.

Meanwhile, I've been using high-volume id'ing techniques to learn about trees in a more northerly region, and wondering yesterday where to collect some sets of obs. Thank you for the generous invitation to use your tree project!

In some regions, identifiers may be using an intentionally much broader id for "duplicates" of other observations they have already id'ed. Some of the sets have identifying features of a tree across several shots. In-app photographers are limited in this way, and may produce large numbers of shots in the field. These sets are still useful. They can be put together later in a by-user (or observer group) time sequence to see features of the same plant. Relevant shots can be annotated. I encourage people to get even the dupes to angiosperm(a) if they can.

Posted by lotteryd 7 months ago

Regarding the family links above, which I've been working through: how do the observations get added to these projects? How are the obserervations removed? Thanks for clarifying. I'm trying to figure out how these will work into a workflow for me. Are they autorefreshing, or does a human have to add observations to them? Thanks to @jeanphilippeb for these

Posted by peakaytea 5 months ago

These are static. Which ones are you working with?
For the families, the family ID is that suggested by the AI CV. However, only those identified to above family are shown in the ID tool. When identified to family (or RG), they are automatically excluded.
I would recommend that if you are systematically working through them, then those you dont want to see again you marked as "reviewed

Posted by tonyrebelo 5 months ago

Thanks. When you say that "these are static," do you mean that no observations are being added?

Posted by peakaytea 5 months ago

They are only periodically added to the projects: hence static. But also, it only changes if you make IDs - if you dont interact then nothing happens.
But it should exclude the stuff you or anyone else has looked at.

Posted by tonyrebelo 5 months ago

Got it!

Posted by peakaytea 5 months ago

Not sure - if JP has abandoned the yellow label family projects - since iNat told him he may not exclude observers who have Opted Out of CID.

He is sorting thru the Pre-Mavericks and moving the done ones to the Archive

Posted by dianastuder 5 months ago

? Its JP's project: he may exclude anyone he likes. Please tell me more!

Posted by tonyrebelo 5 months ago

The iNaturalist Staff asked me to delete the "User has opted-out of Community Taxon (Collection)" project. For simplicity, I did it without discussion. This is motivated by a few complaints about this project from people it lists who feel targeted.
As a technical consequence, I suspend my activity for populating the "Unknown / ..." projects (see: ). I chose not to add "opted-out" observations to these projects and it can't be done efficiently without the "opted-out" project. Until a new feature is available for filtering "opted-out" observations. (A technical review of this feature is on-going).
Posted by jeanphilippeb 6 months ago

Comments on his original journal post

iNat staff prefers to support observers (anonymous) choice to opt out for ALL their obs. But - it is visible (greyed out and off to the side) on their obs. However he may not 'call them out' by putting them in a project so he can exclude them from the projects for which he has, kindly, written the code.

Posted by dianastuder 5 months ago

Surely JP can put them in a table and exclude them without needing a project?
I do refuse to ID observations that are opted out if I notice. I do move them to a higher level if they are wrongly identified.

Posted by tonyrebelo 5 months ago

I won't ID opted outs. And have been snarled at for adding an 'opted out' warning in a comment.

Posted by dianastuder 5 months ago

It would be useful to move any observation to a higher level if wrongly identified. Helping someone get a correct identification isn't a punishment. It's a favor to them.

Posted by peakaytea 5 months ago

Ah, but if you opt out, then you dont want to know if you are wrong - in fact, you cannot be wrong, otherwise why opt out in the first place?

Posted by tonyrebelo 5 months ago

You cannot move an Opted Out anywhere. No matter how many taxon specialists join the queue. That is why I leave a warning.

Posted by dianastuder 5 months ago

I do, too, Diana. Something like "The user has opted out of community ID in their profile." That way the user knows as well. Most of the time, around us, it was an accidental setting change

Posted by peakaytea 5 months ago

If JPs software works by looking at the first photo in the observation, wouldn't it be beneficial to save that slot for the detailed photo?

Posted by peakaytea 5 months ago

No: because often people cannot get detailed photos - e.g. nature reserves and parks where one is not allowed to leave one's car. These can often be identified by subtle jizz, and the AI should be taught how to use that as well. Detailed photos are needed on some observations to anchor the habit shot identifications.
Besides JP may at any moment adjust his software to look at the second, third or even all photos.

Posted by tonyrebelo 5 months ago

Well, that's a twisted bit of logic, but OK

Posted by peakaytea 5 months ago

It is not twisted at all. Just like people can ID plants without flowers or fruit or leaves, so should the AI be able to, and in an ideal world, get it right whenever possible.

Posted by tonyrebelo 5 months ago

You can teach the AI about GISS without having it in the first photo.

Posted by peakaytea 5 months ago

In which case, what is the reason for wanting to have the detailed picture in the first slot?

Posted by tonyrebelo 5 months ago

1) because JP's software would be able to suggest a family 2) because potential expert IDers would see that details are available, rather than glossing over a photo in the gallery of a distant, nondescript bush

Posted by peakaytea 5 months ago

Any expert will trawl through each photo ...

Posted by tonyrebelo 5 months ago

But you don't, and neither do others. It's human nature. Do you want to be effective and get these things IDd, or do you want to maintain some sort of philosophical take?

Posted by peakaytea 5 months ago

Also, the point isn't whether the expert will look through each photo once clicking on the observation. The first goal is getting the expert to click on your observation. Landscape photos are not attractive

Posted by peakaytea 5 months ago

I disagree. Experts dont have time to do identifications for fun or for free. They are trawling for new records and new species and new forms: they will click everything until they know what it is, but only make IDs when they are needed, or when they will help with pushing an ID towards their future use, although they may be using iNat to hone their ID skills.
Time is money and life is short.

Posted by tonyrebelo 5 months ago

Yes, exactly my point! Why would you waste their time by hiding the details! So disrespectful!

Posted by peakaytea 5 months ago

Anyone who is only looking at the first picture - is also ignoring potential placeholder text (and effectively deleting it with their ID)
And also ignoring any notes which the observer may have added (iNat at least doesn't HIDE Notes from subsequent identifiers)

Posted by dianastuder 5 months ago

I disagree. Not disrespectful at all. Looking at closeups of proteas would be boring: its the challenge of various views and features that makes sorting out thousands of proteas interesting.

Posted by tonyrebelo 5 months ago

Diana, the placeholder text shows when you click to ID.
Tony, again, you're talking philosophy, not effectiveness. As you wrote, "Experts dont have time to do identifications for fun or for free. They are trawling for new records and new species and new forms." Are they looking for these by seeing long views of the plant? C'mon!

You will flip your argument from one side to the other to try to justify slowing down the identification of plants by two experts. It is remarkable, and confounding, to see the argument vacillate here.

But my favorite so far is that Tony does it for the amusement of other Protea experts. Nice!

Posted by peakaytea 4 months ago

If you stop misinterpreting what I say, then you wont find any flips, vacillations and confoundings (and no doubt, a different favourite).

Posted by tonyrebelo 4 months ago

Please clarify things for me. I want to understand so that IDs can flow smoothly

Posted by peakaytea 4 months ago

Add a Comment

Sign In or Sign Up to add comments