Taxon swap (lumping), recreation of a lumped name under a different taxon ID, and potential problems

Here’s the story:

Pectiantia was lumped into Mitella via a taxon swap in 2018 in accordance with Plants of the World Online. The name was later recreated in Oct 2022 under a different taxon ID (probably in reaction to this flag). Though evidently not accepted in 2018, Pectiantia is currently accepted on Plants of the World Online. This has the potential to cause confusion and problems, especially if any old identifiers have left. Basically, an ID of Pectiantia was taxon swapped to Mitella and a new identifier IDed it as Pectiantia which disagreed with the taxon swapped ID of Mitella even though there’s no disagreement between the IDs at all. As such, I have 3 questions/comments:

  1. Is there a better procedure to use that would avoid this? Honestly, if I were curating the flag linked above, I would have done the same thing without even realizing there was a previous name.
  2. Is there a way to revert the old taxon swap and then merge the old Pectiantia name with the new one or will this just have to be fixed by manual identification?
  3. If 1) isn’t an option, is there a way to retain old names in a better manner in the case of taxa that commonly get split and lumped over time? It seems like preserving these relationships in some way would make things a lot smoother long-term. Not doing so risks old identifications being changed to the point where they’ve lost meaning over repeated lumpings (this could relate to this post).

FYI, I created a flag for this before I thought it would be better to discuss the problem here.

4 Likes

I don’t think there is any going back and re-activating the inactivated Pectiantia, especially since the circumscriptions could have changed in the interim.

Definitely wouldn’t be able to revert the old swap this far down the road.

I haven’t ever tried, but I’m also guessing there would be no “re-swapping” of the inactivated Pectiantia into the new version. Seems like it would create a data mess with the old taxon swap. But I’d love to be wrong, because otherwise that might fix the disagreement issue.

So yeah, unless I’m missing something, re-identifications may be the only option in the current system.

It’s possible to do a split on Mitella and bump everything which is currently identified at the genus level up to the tribal level see here: https://forum.inaturalist.org/t/new-tool-for-analyzing-unintended-disagreements-caused-by-thinning-parents/31970

It looks like doing so would affect over 500 observations and would lead to quite a lot of reidentification anyway. https://www.inaturalist.org/observations/identify?taxon_id=52105&lrank=genus

If you need votes to sway the community taxon, make an identification link that finds all the problematic observations and I (and other curators) can help bump things to where they need to go.

2 Likes

I was curious to know if it would even let me make a taxon swap with an inactive taxon and it looks like it does. It doesn’t look like it will change any of the names of the occurrences themselves. That said, it might at least list the taxon ID under Pectiantia instead of Mitella. Here’s the link: https://www.inaturalist.org/taxon_changes/120056
I’m kind of curious to know what would happen if I committed it. I can’t really think of anything bad that would happen, but perhaps I’m not that imaginative.

I guess I could, though I honestly know very little about the group and probably wouldn’t feel comfortable doing that without the help of someone who actually knows something about the taxonomic situation.

That would certainly work if I knew much about the group. :-)

I may try adding a comment linking back here to the original flag. Perhaps someone there would like to take a crack at doing a split.

1 Like

Also, moving forward, do either of you have ideas on how we could prevent this situation in the future? As I mentioned above, I probably would have done the same thing, but now that I know the problems it could cause, I’d like to figure out how to modify my process if I can.

1 Like

It would only affect the 2 observations where I assume the observer must have opted out of taxon swap updates.

In terms of preventing this, see the forum link above about the analyze unintended disagreements tool.

2 Likes

The problems are fairly minimal considering the scope of the changes that were put through. To my knowledge, we don’t really have the tools to do a “resurrection” taxon change. One or two identifiers can mostly fix this issue in 15 minutes or less if you identify at the same rate I do. There are currently 76 observations IDed as Pectiantia at tribe level or higher, which you can see find here: https://www.inaturalist.org/observations?ident_taxon_id=1424152&lrank=tribe&place_id=any&quality_grade=needs_id&subview=map

Here’s the Identify link: https://www.inaturalist.org/observations/identify?verifiable=true&page=1&spam=false&place_id=any&ident_taxon_id=1424152&user_id=&project_id=&quality_grade=needs_id&lrank=tribe&per_page=100

2 Likes

Without some input from @loarie I’m not feeling as confident about that. Don’t feel like I have a complete picture of the ID data relationships behind the scenes.

The old taxon swap still exists in the system, but I don’t know if it continues to be referenced when activity feeds on an observation are constructed, etc. There would be two different swaps in the system referencing the same input taxon. I suppose the upshot might be the equivalent of a “split” of the input taxon, with 100% going one way and 0% going the other. But again, without a fuller picture of data structures and interactions, I’m not sure.

The two observations showing in the draft swap actually would not be affected, since the observer continues to opt out of taxon changes.

But I wonder how many other hundreds of observations might be affected that aren’t included in that count. I don’t think a taxon swap involving an inactive input taxon would act on old inactive IDs (withdrawn by the system due to the previous taxon change) - but I’m not sure there have been any previous test cases that would verify that.

If the problem is

I would say that there actually is a disagreement there, unless Mitella and the new Pectiantia have exactly the same circumscriptions, in which case one would just swap Mitella into Pectiantia to cure the disagreements. But if only a part of Mitella belongs back in Pectiantia, then to remedy the disagreements it sounds like Mitella needs to be (re-)split, and the appropriate species swapped to equivalent Pectiantia species.

That seems too simple, so I’m probably missing the crux of the problem here… :thinking:

I think this kind of problem is, in effect, a deliberate feature of how iNaturalist handles taxonomy and identifications.

Consider two alternative models for handling taxonomic change.

Model A:

name 1 → taxonomic change 1 → name 2

name 2 → taxonomic change 2 → name 3

Model B:

name 1 → taxonomic change 1 → name 2

name 1 → taxonomic change 2 → name 3

We can think of model A as a chain. Each taxonomic change acts on the output of the previous taxonomic change. In model B, each taxonomic change acts on the first name given to the observation. The situation with Mitella and Pectiantia is a basic feature of how the model A / chain handling of taxonomy functions. Possible solutions involve finding a way to “trick” the chain into not acting like a chain…

1 Like

(More broadly, the model A / chain approach has inherent problems with tolerance stacking—error rates in taxonomic translation are going to stack across the set of taxonomic changes. Not a big deal for a young system that hasn’t accumulated many changes, but a problem that we should expect to become much worse as the system ages…)

2 Likes

I believe the best practice when splitting a name on iNat is to create a taxon split for the (for example) genus. If there’s a geographic split between the two (or more) taxa, the split should be atlased so that observations at the genus level can be sorted into the proper new genera. If atlasing is not possible, then they get bumped back up to the next taxonomic level, and then will have to manually be reconfirmed as one genus or the other. This re-examination should happen regardless, but if they are all bumped up, it eliminates the problematic disagreement mentioned in the original post.

Of the changes that occurred, none were a split, in the iNat sense. It was a taxonomic split of Mitella that resulted in observations being raised to tribe level that were identified as both genus Mitella and a species moved out of Mitella. There are still some observations identified as Mitella s.l. outside the range of Mitella s.s., as well as the aforementioned observations. It’s complicated, but what I believe Nathan is asking is whether or not “resurrected” taxa, like the change mentioned above, should or could be handled in a more neat manner.

Nathan, I feel your pain. There was a recent lumping of a couple of moth genera which was inappropriate and we’re finding out that (a) staff cannot summarily undo a taxon swap/lump, and (b) if the genera were to be resplit, it generates a new taxon ID for at least one (or both?) of the previous genera. This involves some 75+ species and nearly 7,000 observations. Such cases provide a strong argument for letting any proposed taxon swap linger and simmer for a few to several days or weeks to garner comments, rather then committing it on the same day it was proposed. I wish there was an automatic, built-in waiting period for any proposed swap. I don’t know if that has been proposed as a feature request.
EDIT: I’ve submitted a Feature Request for a waiting period for the committing of taxon swaps after a proposal. Watch for that.

2 Likes

Good to know we can expect data loss in our future (ugh…). I hope this gets addressed at some point.

I guess my thoughts are more towards identifications made by iNatters that have gone inactive or those who don’t understand the situation. It may only take 15 minutes to ID, but could be followed up by hours of writing to confused observers. Or worse yet, the observers just keep their own IDs. Then the inactive identifiers and the observers who don’t change will have to be manually overruled by at least two other identifiers. If that ID is found to be wrong in a year or two, all users must change or get even more identifiers involved. Another taxon change happens and you see the problem. At some point, observations aren’t worth saving (or even if they are worth saving, the time and energy involved in fixing them is just too much). Not suggesting it will happen here but the irreversibility issue is kind of starting to worry me.

Also, I want to be clear that I have no criticism with the taxon swap itself. I don’t really have enough time to learn the group enough to criticize it even if I wanted to and this looks like a complicated one. :-) I just want to figure out if there’s a better way to do it with the system we’ve got.

Actually, maybe that’s a good argument for making some disagreeing identifications going inactive after some period of time (a topic that I could be sure was discussed before, but I can’t find it), but then you have the data loss problem again.