Taxonomic splits with atlases: how much overlap is okay?

When making an atlas-informed taxonomic split, how much geographic overlap in the ranges of the destination-taxa is okay? Should I remove much of this overlap by first constructing finer-resolution atlases of the destination taxa?

I’m sorry for the long post, but I think this situation must be somewhat common, if a little complicated.

I’m drafting a taxon split to bring our taxonomy of a species in-line with POWO (the authority we use for vascular plants).

The taxon being split is a plant species with a large distribution across North America north of Mexico. My drafted change splits it into multiple new species, which align with taxa formerly long-recognized as subspecies or varieties. The new, narrower species have ranges that overlap considerably on the state- and province-level. More than half of the observations of the old, widely-distributed species occur in states and provinces of overlap. However, on a finer geographic scale the new species ranges overlap much less; the new species are mostly, but not completely, allopatric. Just not at the state-/province-level

I have constructed atlases of the destination taxa based on maps drawn by the author of the source ultimately cited by POWO. The source maps depict ranges as shaded regions overlaying political boundaries. These maps show state- and province-level boundaries, but not county- (or equivalent) level boundaries. They are fairly high in resolution as far as voucher-informed range maps usually go, and county-level boundaries could be reasonably guessed-at from the shaded areas. So far, my atlases are only at the state- and province-level. It would be a significant amount of work to translate the source maps into higher resolution iNat maps, but not impossible.

If I continue with this taxonomic split, my understanding is that observations of the old taxon in the states and provinces of overlap will be bumped up to the next-higher taxon (in this case, only up to subsection, not genus, so not the end of the world) and the observations in the states and provinces of no overlap will be moved to the new taxa there. Since there is so much overlap, this could be confusing and disruptive. But the status quo is also not great. What should I do?

The proposed splits are here:
https://www.inaturalist.org/taxon_changes/86155
https://www.inaturalist.org/taxon_changes/86154

A better place to leave comments specific to the details of these taxa:
https://www.inaturalist.org/flags/501223

1 Like

Sort-of related:
https://forum.inaturalist.org/t/whats-the-right-way-to-do-this-hypothetical-taxon-split/4414

For https://www.inaturalist.org/taxon_changes/86155 I recommend the new (new as of September) approach of using the taxon split input as an output - ie Solidago speciosa 130376) - especially if Solidago speciosa is allopatric to the other species.

If you can make the output species allopatric to one another by exploding the atlases down to counties I’d recommend doing that

to the degree that output species atlases overlap, as you say, this split is just going to replace IDs of Solidago speciosa with IDs of the subsection.

It would be good to ask which is more annoying - having a bunch of species level IDs coarsen to the subsection, or not do the split and let this play out in the ID realm (ie “thats not Solidago speciosa its Solidago rigidiuscula”). In my opinion splits are definitely worth it when what we mean by a species is well known by the community (ie the single moose that occurs in both America and Eurasia) and we want to change that. But in cases where the set of species and the boundaries between them are poorly known and ambiguous / misunderstood among the majority of the community to begin with, it might be more direct to shore things up via the ID process itself by adding IDs.

2 Likes

So in letting this play out in the ID realm, you’re envisioning something like:
“I think this is actually species Y, which some authors recently moved out of species X. This change is not currently represented in the taxonomic framework iNaturalist uses, which at the moment shows Y as a sub-entity of X, but it is present in the source iNaturalist uses.”

The first question is do we want to follow POWO or rather deviate. I assume you want to follow POWO?
If so then you need to make sure Solidago pallida, Solidago jejunifolia, and Solidago rigidiuscula are active alongside Solidago speciosa.

The next question is what to do with existing IDs of Solidago speciosa that may not match POWO"s narrowed interpretation. Your two options are:

  1. let it play out in the ID realm:

pros: won’t coarsen lots of observations by replacing IDs of Solidago speciosa with IDs of the subsection, won’t incur the infrastructure and notification costs off all the notifications and IDs generated by the split

cons: the system will interpret any existing IDs of Solidago speciosa as disagreements with other taxa like Solidago pallida which may make it harder to properly ID these observations. Also, it may make some identifers look like they have alot of wrong IDs when really their ID was right, we just changed the meaning of it.

  1. commit the split with Solidago speciosa as the input and Solidago speciosa, Solidago pallida, Solidago jejunifolia, and Solidago rigidiuscula as the outputs

pros: will eliminate all existing IDs of Solidago speciosa so no risk of misinterpreting them as disagreements with species like Solidago pallida

cons: will coarsen a lot of observations (especially if there aren’t non overlapping atlases) and will generate alot of notifications and identifications and delayed jobs which may bog down the system for a bit

To decide which path to take: split or let things play out in the ID realm, its useful to know how well understood the taxa in question are among the iNat community. For a clearcut example, if we were talking about calling all Bobcats west of the Mississippi a new species ‘western bobcat’ then this is a clear example where Bobcat needs to be split (bobcat as the input and bobcat and western bobcat as outputs) because what we mean up to this point by Bobcat is well known and understood by the community. So if we suddenly narrowed what we mean by Bobcat to mean just observations east of the Mississippi, all those existing IDs of Bobcat that meant Bobcat in the broad sense but now are being interpreted as Eastern Bobcat in the narrow sense disagreeing with Western Bobcat would be a big problem. A split helps notify everyone that we’ve changed what we mean and removes now ambiguous IDs (e.g. Bobcat)

Alternatively, if we’re talking about a genus of fungi where its unclear how many species there are and looking at the existing observations on iNat there seems to be a lot of confusion about what they mean, how to tell them apart and where they occur, and I wanted to help clear things up around these taxa and part of that meant making it clear that Mushroom A is distinct from Mushroom B, its probably not worth splitting Mushroom B and rather helping this play out via the ID realm by posting IDs of Mushroom A along with comments on observations that also have IDs of Mushroom B explaining how these species are different and why the observation is likely Mushroom A and not B.

These are extreme examples, its not clearcut in the gray area in between. I don’t have a good sense for Solidago taxonomy so I’m not sure where this case falls on that gradient. Does this help at all?

1 Like

I think we’re talking past each other a little bit. One of my motivations for splitting the taxon to begin with is to be able to have those ID realm discussions. I think it’s difficult to have them when there are multiple iNat entities representing, say, Mushrooms A and B at various taxonomic levels. Don’t we have to sort out the taxonomy first, so people have some idea of what’s even being suggested when someone suggests an ID on an observation? As it stands in my example, there are actively-used entities at the levels of variety, subspecies, and species.

As for this particular case: I’m very open to explicitly deviating from POWO. I’m basically agnostic on the judgment call, although I was hoping to have discussions of the details of this particular case on the flag comments. The inactivated taxa you mention are inactivated because that was the only way I could put them into a species-complex with the same name as an existing species, without making any changes to taxa in current use. The system does not allow for active species and species complexes with the same name at the same rank.

2 Likes

apologies if we’re not understanding one another, this stuff is complicated. I’m not sure if this will help or further confusing things, but this is how I would go about cleaning up Solidago and how that process leads into the questions about splitting.

Step 1: decide whether the goal is to curate in the direction of POWO or to deviate from POWO. Lets assume the goal is to curate in the direction of POWO since that’s definitely simplest

Step 2: handle taxa in iNat that are not in POWO

Step2a: internodes
POWO only has families, genera, species, and ssp/var so it doesn’t offer guidance on things like sections. I assume we want to keep them though so I accommodated them with a deviation
https://www.inaturalist.org/taxon_framework_relationships/464705
and included a link to a flag I made where the rationale behind these nodes and which taxa descend from them can be discussed (since we can’t refer to POWO here). There are also quite a few species not assigned to section or subsection that should be moved if possible (may not be possible).

Step2b: ‘relationship unknowns’
There are 55 Solidago in iNat that aren’t in POWO (and aren’t accounted for with deviations) we call these ‘relationship unknowns’. Assuming we’re not deviating, they can usually be dealt with by swapping them into other existing taxa (or in some cases new taxa that don’t yet exist that you’ll need to make). I usually deal with these first. (one caveat is that for some reason POWO often is lacking the nominate infraspecies e.g. Solidago curtisii var. curtisii - since we want these to exist on iNat, rather than swap these away its better to keep them and create not-external deviations to account for them)

Solidago araucana
Solidago arguta var. arguta
Solidago caesia var. caesia
Solidago canadensis var. canadensis
Solidago chilensis var. chilensis
Solidago curtisii var. curtisii
Solidago elongata
Solidago fallax var. fallax
Solidago flaccidifolia
Solidago gigantea var. gigantea
Solidago gigantea var. leiophylla
Solidago hispida var. arnoglossa
Solidago hispida var. hispida
Solidago hispida var. huronensis
Solidago hispida var. lanata
Solidago hispida var. tonsa
Solidago hybrida
Solidago lepida ssp. lepida
Solidago lepida
Solidago maheuxii
Solidago microglossa
Solidago missouriensis var. fasciculata
Solidago missouriensis var. missouriensis
Solidago missouriensis var. tenuissima
Solidago missouriensis var. tolmieana
Solidago mollis var. angustata
Solidago mollis var. mollis
Solidago multiradiata var. scopulorum
Solidago nemoralis var. longipetiolata
Solidago petiolaris var. angusta
Solidago petiolaris var. petiolaris
Solidago porteri
Solidago radula var. laeta
Solidago radula var. radula
Solidago radula var. stenolepis
Solidago rugosa var. rugosa
Solidago simplex var. ontarioensis
Solidago simplex ssp. simplex
Solidago simulans
Solidago spathulata var. gillmanii
Solidago spathulata ssp. glutinosa
Solidago spathulata var. nana
Solidago spathulata ssp. spathulata
Solidago speciosa var. jejunifolia
Solidago speciosa var. rigidiuscula
Solidago speciosa ssp. speciosa
Solidago uliginosa var. levipes
Solidago uliginosa var. linoides
Solidago uliginosa var. peracuta
Solidago uliginosa var. terrae-novae
Solidago uliginosa var. uliginosa
Solidago ulmifolia var. microphylla
Solidago ulmifolia var. ulmifolia
Solidago virgaurea var. praeflorens
Solidago wrightii var. adenophora
Solidago wrightii var. wrightii
Solidago × asperula
Solidago × bartramiana
Solidago × calcicola
Solidago × luteus
Solidago × neurolepis
Solidago × ovata
Solidago × raymondii
Solidago × ulmicaesia

Step 3. taxa in POWO not in iNat
there are 30 Solidago in POWO not in iNat. Assuming we’re not deviating and the goal is to make Solidago complete (ie add all the taxa to iNat in this clade), these need to be created.

Solidago argentinensis
Solidago calcicola
Solidago canadensis var. lepida
Solidago canadensis var. scabra
Solidago compacta
Solidago correllii
Solidago jejunifolia
Solidago kuhistanica
Solidago multiradiata ssp. paramuschirensis
Solidago nipponica
Solidago ontarioensis
Solidago ovata
Solidago patagonica
Solidago patuliginosa
Solidago polyglossa
Solidago rigidiuscula
Solidago spellenbergii
Solidago ulmicaesia
Solidago velutina ssp. velutina
Solidago virgaurea ssp. armena
Solidago virgaurea ssp. gigantea
Solidago virgaurea var. insularis
Solidago virgaurea ssp. jailarum
Solidago virgaurea ssp. kurilensis
Solidago virgaurea ssp. pineticola
Solidago virgaurea ssp. talyschensis
Solidago virgaurea ssp. turfosa
Solidago yokusaiana
Solidago × leiophallax
Solidago × maheuxii

Steps 4. Do we need to split anything?
Now the structure of the tree in iNat should match what we’re aiming for -that is POWO plus the deviation to accomodate the sections etc. But there’s still this question about whether adding any of the new species narrow what we meant by any of existing species and thus change the intention of existing IDs such that we need to split some of those existing taxa to replace those now ambiguous problematic existing IDs? If so we need to split.

For example, and back to your original example, did the fact that we added Solidago jejunifolia etc change the meaning of a lot of existing IDs of Solidago speciosa such that we should probably create a split like https://www.inaturalist.org/taxon_changes/86155?

Sorry if this just continues talking past one another. But that describes how I would go about cleaning up Solidago. If you can give another go to help me understand your issue, I’ll try again to see if I understand the problem and can help (or maybe better to be quite and let others help!)

I believe that atlases can be made up of a mix of different levels of geographical detail, so you could create an atlas for each target species that includes:

  1. Those whole states and provinces that are not within the distribution of any of your other target species.
  2. The specific counties (etc.) where more than one species is present in the target state or province.

Would that reduce the atlas work required to a manageable level?

I do think it should be acceptable for you as a curator to use your judgment to reflect the source maps’ shading in selecting these counties (if you choose to take on that work).

1 Like

For me personally, I often make new range maps in QGIS and import them into iNat. Once the map is imported, you don’t have to guess at the county-level, it’s overlaid on the atlas map and you just turn on the ones that have some portion of the range in them. I will admit it can be some extra work, but with everything already set up and ready, I would say usually less than 20 mins per map.
You can also try contacting the author and asking for their map files, though I’ve had only mixed success with this route.
If you want a bit of a shortcut, you can just georeference the source maps onto county-level maps and skip the iNat upload. Here’s one for Solidaga pallida I did in about 5 minutes (county map from GADM):

As has been said above, I think there’s a balance between atlasing fine and coarse. If it’s very coarse, you potentially lose information and make identifiers upset that you haven’t captured their intent, and if it’s very fine I think you risk misrepresenting the certainty of IDs based on location alone (keeping in mind possible errors both in the range maps and in the iNat observation locations).

3 Likes

Yup, that’s about what I was going to say. :-)

I hope some day people will just upload their GIS data to one of the various data repositories–presumably Semple has a shapefile, kmz, or the like, and it’d save us having to recreate them! OTOH, for all we know these could be hand-drawn in an image editor, in which case a faithful reproduction of the polygons is likely to faithfully reproduce errors.

2 Likes

I’m not going to name any names, but I know a big taxonomic source that indeed created all its maps in Adobe Illustrator. :woman_facepalming:

3 Likes

There’s one taxonomist who provides me with some minor amusement by publishing dot maps using an obsolete map of county boundaries that was partially corrected to the current boundaries, so that it doesn’t quite match the county boundaries at any particular point in time.

3 Likes