There's a draft taxon change here https://www.inaturalist.org/taxon_changes/89068 that may be unnecessary or even undesirable. (note: the output taxa used to be grafted to a new C. arvensis, currently inactive)

Castilleja arvensis var. pastorei was named in 2018 and added to iNaturalist in 2019. I'm not sure what its taxonomic history is on Plants of the World Online, but it is currently accepted there. New IDs of these varieties can be added by people interested in IDing to variety. I can't think of a similar example of splitting a species into its infrataxa. I don't think we do that, do we @loarie? We just use manual IDs for that?

cc: @jdmore @markegger

Copying comments from concurrent discussion:

jdmore: Platform (Android, iOS, Website): web site
Browser, if a website issue (Firefox, Chrome, etc) : Firefox 84.0.1
URLs (aka web addresses) of any relevant observations or pages:
https://www.inaturalist.org/taxon_changes/89068
[EDIT - taxon change now deleted, see discussion below and at https://www.inaturalist.org/flags/521779 for more context]
Description of problem (please provide a set of steps we can use to replicate the issue, and make as many as you need.):
Step 1: Attempt to commit the above atlased taxon split
Step 2: Commit fails, leads to “I got nuthin” (404 Not Found) page
Can’t see anything wrong with the split set-up, but maybe I’m missing something…

bouteloua: If you want to add subspecies, we just add them without any taxon changes and people can add refining IDs if they want. A taxon change like this inactivates the input, which we wouldn’t want, and is probably why it’s breaking. I reactivated the input.

jdmore: So it was set up like this:
INPUT:
C. arvensis (133740 - active)
OUTPUT:
C. arvensis pastorei (inactive, parented to C. arvensis 1197997 inactive)
C. arvensis arvensis (inactive, parented to C. arvensis 1197997 inactive)
I wonder if the problem was that I did not explicitly include the inactive parent (C. arvensis 1197997) as one of the output taxa?
Since the species has about 1000 observations, and the varieties are well-defined geographically, the hope was that doing a split would save the Castilleja expert some labor re-identifying them all.

jdmore: Also, note that the C. arvensis pastorei that you activated was an earlier version inactivated via a previous taxon change. So it’s the new C. arvensis pastorei that should probably be activated, right?

bouteloua: The old taxon still had several active IDs somehow (Mark opting out of taxon changes?), so my thought was that it should be prioritized.

loarie: We should probably add a validation where a taxon split can’t be committed if all output taxa are descendants of the input. Would that be the best course of action here? Would be great to find a way to make it clearer that taxon changes shouldn’t be used to refine IDs (e.g. species → ssp)

bouteloua: I mussed it all up, confusing things. During my comment I hadn’t realized until looking into it further that the output taxa actually had a different parent (inactive, https://www.inaturalist.org/taxa/1197997-Castilleja-arvensis) than the input.

jdmore: So if a species previously had no subspecies, and then geographic subspecies were subsequently named / described / recognized, that would not be a legitimate use of a taxon split? If so, I guess that wasn’t sufficiently clear to me before. Only “horizontal” splits and no “vertical” splits, so to speak?

loarie: correct - if a species was previously monotypic (e.g. no ssp) and then later on ssp A and ssp B were subsequently recognized, just add the new ssp descending from the parent species - no taxon change needed. The community will have to help push observations ID’d to species forward to ssp with new IDs

Say C. arvensis var. pastorei was described in 2021 and "new" to the taxonomic tree on iNaturalist though, somewhat analogous to a species being split into two or more species.

In that case, perhaps a taxon change could be "used to make sure content gets moved to the right spot in response to alterations to the structure of the tree" (-loarie in those comments) ? Or are infrataxa just a wholly separate consideration than taxa at species or coarser?

@loarie that active C. arvensis pastorei had been inactivated last year by a different taxon change, so I think it is best left inactive to maintain the integrity of that previous taxon change, and activate a new var. pastorei instead for this newest change. (The 3 remaining IDs using the "old" pastorei are by @markegger, who presumably has opted out of automatically accepting taxon changes -- I will communicate with him about adding "new" pastorei IDs to those observations.)

All - as the now-better-educated instigator of the above taxon change, and with the above said, I have cleaned up the mess as follows:

1- re-inactivated the old https://www.inaturalist.org/taxa/882518-Castilleja-arvensis-pastorei
2- activated the new https://www.inaturalist.org/taxa/1196830-Castilleja-arvensis-pastorei and grafted it to the old https://www.inaturalist.org/taxa/133740-Castilleja-arvensis
3- deleted the new https://www.inaturalist.org/taxa/1197997-Castilleja-arvensis, now rendered unnecessary
4- updated the Taxon Framework Relationship for C. arvensis and varieties: https://www.inaturalist.org/taxon_framework_relationships/469734
5- deleted the above-referenced taxon split since it is no longer necessary.

Again, I will contact Mark about dealing with the old pastorei IDs.

The biggest concern I would have about using taxa splits to introduce newly described subspecies is questioning if the range & overlap is really worked out at that point. I suspect few studies that describe a new subspecies will come with enough data to be confident in this biogeographic data.

Surely, assuming there are clear visual keys that can be used to separate the different items, it is better to use human eyes to do the ID's which in turn helps to solidify that range data. As with all ssp id's based on range data, I fear any automated carving out of new ssp's just becomes circular reasoning, the paper says this is here, so that must be what it is.

It's odd to me that a record existed for var. pastorei from 2019, as I've tried to use this name unsuccessfully several times since then. BTW, this is not a "new" var., as it was described decades ago but languished since it was assigned to the wrong species. My 2019 paper was a comb. nov., moving it to C. arvensis. Its range is essentially allopatric with the nominate form, except in eastern Brazil, where they become parapatric. This is confirmed both by herbarium records and iNat sightings I've reviewed. I had hoped to accomplish the split of the 1000+ existing records without having to hand-annotate all of them, except in the zones of potential co-occurrence, which we intentionally left out of the areas defined for the split.

@jdmore splits should only be used when 'what we mean' by a taxon has narrowed rendering existing IDs that used to be agreements as disagreements. For example if Species B is carved off of Species A, then existing IDs of Species A made when it was implied to be broader/sensu-lato become interpreted as the narrower/sensu-stricto Species A and thus disagreements with Species B. Making a split solves this problem by replacing existing IDs of Species A with a common ancestor of Species A and B (ie not a disagreement) or if we can use geography (e.g. atlases) replaced with IDs of Species B where appropriate.

In the vertical split situation you describe existing IDs are never re-interpreted as disagreements so a split isn't necessary. Apologies this stuff is so confusing - we really need better documentation on taxon changes.

Initially, splits would identifications of Species A with identifications of the common ancestor of Species A and Species B. There was a lot of legitimate concern about unnecessarily coarsening of identifications - especially when the carved off Species B represented a small geographically distinct population.

The atlas functionality was meant to address data loss/coarsening concern by allowing people to split off Species B while avoiding replacing IDs of Species A with IDs of the common ancestor (instead based on geography/atlases IDs are replaced with IDs of Species B or left as Species A).

My main concern with using taxon splits + atlases to replace IDs of Species A with IDs of Subspecies B is that there's no coarsening/data loss issue that it would alleviate. And it opens a very slippery slope that @cmcheatle mentioned of people using taxon changes to 'bulk ID' which is problematic because it subverts the ID process on iNat.

I see your point that even using a taxon split + atlas to split Species B off from Species A kind of subverts this process by replacing some IDs of Species A with IDs of Species A based on geography alone. If people feel like thats a legitimate subversion of the ID process that should be reigned in my preference would be to alter that functionality so that atlases are used to leave IDs as Species A OR roll the ID back to the common ancestor (of Species A & B). Or remove the taxon split + atlas functionality altogether or otherwise restrict its usage so that its much more common that the carving Species B off from Species A will roll ID's back to the common ancestor (and incur the data coarsening/loss hit). Curious what others think

I must say, I agree with Jim's perspective on this, point after point. Of course, he's supporting what I'd hoped to achieve, so maybe that's bias on my part. But remember in this case, this is a sub-specific taxon, validly published decades ago and only neglected because it was associated with the wrong species at the time of publication and was in a region rather neglected in the literature. AND it has a nearly parapatric range from the nominate form. I've already annotated all currently posted var. pastorei and all the var. arvensis records from adjacent regions of southern South America, so in that sense the split is accomplished. But this IS a case where the two forms are quite readily separated by both range and an easily identified field character. They apparently overlap only in a small region near Sao Paulo, Brazil, though the occurrence of this species in Brazil is vastly underrepresented on INat at the moment, I believe. It will be interesting (at least to me!) to see how the future post from Brazil fall out.