Replace hybrid ranks with a hybrid_level property

Feature Requests
1

Platform(s), such as mobile, website, API, other: All

URLs (aka web addresses) of any pages, if relevant:

https://forum.inaturalist.org/t/species-search-doesnt-include-intergrade-observations/57768

Description of need:
The way hybrids are currently implemented is lacking. Because there are separate ranks for hybrids, hybrids can only be added for those ranks which have a hybrid counterpart (genus, species, subspecies, variety and form). This prevents things like family hybrids (like Icteridae Ă— Icteriidae) and section hybrids (like Rubus Ă— Ursini) from being added. And the search cannot distinguish between hybrids of different levels, such as an intraspecific hybrid between subspecies (most infrahybrids) and an interspecific hybrid between subspecies (these ones). For the issues with this, see the above bug report.

Feature request details:
Any taxon of the rank “hybrid” should be automatically changed to “species”, and its hybrid_level set to 10 (the rank_level of species and hybrid), though this can be changed manually if necessary.

Any taxon of the rank “genushybrid” should be automatically changed to “genus”, and its hybrid_level set to 20 (the rank_level of genus and genushybrid), though this can be changed manually if necessary.

Any taxon of the rank “infrahybrid” should be automatically changed to “subspecies”, and its hybrid_level set to 5 (the rank_level of subspecies, variety, form and infrahybrid), though both of these can be changed manually if necessary. For example, Anser anser domesticus × cygnoides domesticus will need a rank of variety and a hybrid_level of 10.

Non-hybrids should have their hybrid_level set to 0 (not currently used as a rank_level, and won’t be able to be used as one in the future if this is implemented).

Hybrids between ranks that previously weren’t possible should be added.

Species search doesn't include intergrade observations
2

I feel like the ideal situation would be that when you create a new hybrid, there are 2(+) taxon fields where you can select any taxa. Once you choose one, the other(s) must be of the same rank.

3

Unfortunately this does not work well for plants, where some groups have infraspecific Ă— species combinations, e.g. Asplenium Ă— inexpectatum (A. rhizophyllum Ă— ruta-muraria var. cryptolepis) or more complex nothospecies situations like Carex Ă— flavicans, where nothovar. flavicans is C. aquatilis var. aquatilis Ă— subspathacea and nothovar. substans is C. aquatilis var. minor Ă— subspathacea.

2 Likes
4

Now that I think about it, it doesn’t technically work for other taxa either. For example you could have a hybrid with Canada Goose and a domestic goose where you can’t tell if both Greylag and Swan Goose are involved (i.e. Branta canadensis × Anser). However even though that option could theoretically exist now as far as I know, it doesn’t and so that example would be forced to Branta × Anser.

Looking at the species page for Asplenium Ă— inexpectatum, it looks like the hybrid formula is presented as Asplenium rhizophyllum Ă— ruta-muraria so it seems like maybe the same curation rules are being used for plants as well?

5

I think the person who added the hybrid formula name just didn’t specify the variety. I edited the hybrid formula.

We try to be as specific as possible with plants because there is a tendency for infraspecifics to gain momentum into becoming full species. I’m not familiar with your goose example so I can’t say how it compares, but hybrids in plants involving infraspecifics tend to be more significant than one might expect due to the aforementioned momentum. If you want to see some funny hybrid formulae, search Asplenium trichomanes, which contains a bunch of cryptic speciated lineages that all hybridize with each other and most other sympatric Asplenium. 40 taxa for the price of one species search!

1 Like
6

Interesting, are there different hybrid names for hybrids of A. rhizophyllum with different varieties of ruta-muraria? From my zoology perspective it seems strange to formalize that information as part of the nothospecies name definition rather than having it at a lower rank below the hybrid or even just in publications associated with the hybrid name.

I’m curious if there are any examples of animals where multiple distinctive subspecies of one species hybridize with a second monotypic species. The closest example I can think of was the observation description of this observation which refers to Junco phaeonotus × J. hyemalis dorsalis, but technically J. p. palliatus × J. h. dorsalis would be more precise. Anyone using scientific names would usually just say J. phaeonotus × J. hyemalis rather than including subspecies since there’s only one subspecies of each species breeding where their ranges overlap as far as I know.

7

That was a previous feature request of mine.

https://forum.inaturalist.org/t/list-parents-of-hybrids/46968

P.S. And I’d say there are some situations in which you’d want to have mixed-rank hybrids, like Setophaga coronata auduboni × setophaga townsendi, which I wish hadn’t been deactivated. But the reason a mixed-rank hybrid is warranted in this case is due to Setophaga townsendi being monotypic. I suppose such cases could be exempted from the general rule that mixed-rank hybrids aren’t allowed. But it seems chaotic. What happens if someone adds subspecies to such a monotypic species?

Anyway, I don’t think having parents of hybrids listed should be the only way for iNaturalist to know something’s a hybrid. In the beginning, the parents will probably have to be added manually for at least some hybrids.

1 Like
8

Aha thank you for this warbler example. And from the flag discussion on it, yes, the Curator Guide says that hybrids should be between taxa of the same rank.

1 Like
9

The Curator Guide can be revised if the framework regarding hybrids is revised.

I do like the idea of allowing a user to select the taxa of the parents, but at some point we do have to consider organisms that:

  • do horizontal gene transfer (usually asexual organisms)
  • can have more than two biological parents
  • are hybrids of fertile hybrids (say a coywolf that is a cross between a “purebred” Canis latrans and “purebred” wild Canis lupus has a litter with a domestic dog…totally possible in the United States if the coywolf doesn’t eat the dog)
  • are a part of a ring species or something where it’s common for every organism to be a “hybrid” of “hybrids” of “hybrids” that are still fertile, such as certain salamanders
  • are polyploid and can have more than one possible option of the number of sets of chromosomes they receive from parents, which may affect how biologists classify a hybrid since some have more of parent A than parent B
  • (extremely hypothetical) are offspring of superfecund parents where some gene transfer between developing fraternal twins where at least one twin comes from a father of a different species. This is definitely not something that can be done on iNat unless someone genetically tests two members of the same litter. Most people who genetically test their observations are in microbiology or mycology and not general veterinary science where this is probably most likely to occur and be observable. However, it is important to consider
  • people’s familiarity with how hybrids work on iNaturalist, with some users reluctant to follow new best practice

I can’t make any complaints about how such a revision to the hybrid framework would affect the computer vision model because it doesn’t work well for a lot of hybrids anyway. The best I’ve seen it do is identify hybrids in the birds of Baeolophus.

The only intuitive way I could see this implemented is if every ID was of the taxa of an organism’s parents, where non-hybrids have redundant values for each parent because they are of identical kind. It is also extremely unlikely that we can actually do anything about implementing this framework since there are millions upon millions of existing iNaturalist observations, converting the data structure used to hold taxon IDs would take a long time and possibly add significant bloat to the size of the databases.

If iNaturalist decides to make the father’s ID come before the mother’s ID as is typical when trying to describe the exact parentage of a hybrid (think mules vs. hinnies), then there would be a separation between the correct taxon ID for a mule (Equus africanus asinus × Equus ferus caballus) than a hinny (Equus ferus caballus × Equus africanus asinus) that didn’t exist before. If this is seen as a problem, especially if it’s too much work for something that doesn’t really matter, they can be lumped together as the same taxon and the order of parents would not matter.

Regarding the original post up to and excluding inter-rank hybrids, I like the idea overall. This will add a column or two to some table somewhere in the database, but I can’t predict how much of a hassle it would be to have to expand storage to fit all preexisting observations with a hybrid level. It will also require a whole lot of taxon flags because people with very specific pet taxa will want to flag and add their even more granular hybrid where a more general hybrid exists and is not of great biological importance.

If someone can accurately estimate how much a hybrid_level property would cost to the back end, please reply because my vote depends on this idea being implemented efficiently.

10

Another thing I’d like to add is that I also want the rank_level of variety lowered (perhaps to 3 4), and the rank_level of form set to an even lower number (maybe 2), but of course, both should remain positive. Only subspecies should get to keep the 5.

I want this done at the same time as the implementation of associating hybrids with their parents and the addition of the hybrid_level propery, and the removal of the hybrid-specific ranks. By the way, I’d like the system to automatically assign hybrids to parents based on their scientific names or hybrid formulae. If an infrahybrid has varieties or forms assigned to it as parents, its rank should be set to that rank, rather than subspecies. The lowest rank prevails, which is why variety must be recognised as below subspecies and form must be recognised as below variety. Any hybrid which the system could not assign to parents should be automatically flagged, as it probably needs curator attention.

The hybrid_level property should be set according to rank_level at which the parents diverge. If the parents have been set, it should be possible to do this automatically.

When a curator adds parents to a hybrid, the system should require the parents to be the same rank as the hybrid, with an exception for eligible taxa of a higher rank. This exception only applies if the hybrid is below species. The excepted higher taxon must be species or below, and must not have any children at or above the rank of the hybrid. In other words, an eligible taxon cannot contain eligible taxa. However, the parents can’t all be eligible higher taxa. At least one must be the same rank as the hybrid. There’s no such thing as eligible lower taxa. The hybrid’s rank must be the same as its lowest parent.

With the lowering of the rank_levels for variety and form, and the removal of genushybrid, hybrid and infrahybrid, there would be no ranks with the same rank_level.

By the way, some plant species don’t have subspecies, only varieties. Under my proposal, these species would be allowed hybrids with subspecies, because they are monotypic as they don’t have subspecies, and variety is below, not equal, to subspecies.

P.S. Variety should be rank_level 4, not 3, because of the possibility of adding subvariety.

P.P.S. https://forum.inaturalist.org/t/adjust-rank-hierarchy-so-subspecies-variety-and-forma-are-not-at-same-level/60159