I’ve encountered similar choices about whether to subdivide plant genera. To be clear, I’m talking about subgenera, sections, etc. that have validly published names. As @upupa-epops makes clear, POWO is not going to come to our rescue here as they don’t concern themselves with ranks between genus and species. Here’s what I have done:
- After quite a while working with Sisyrinchium, I decided that it was worth my effort to implement the 10 subgeneric sections outlined in Inacio et al (2017). Some factors in favor here were based on utility for iNat users: It’s a fairly large genus (~200 species), with some easy to identify and others very hard to distinguish, meaning that a lot of observations can be placed confidently in a section but not refined to species. The quality and likely durability of Inacio’s paper was also a big factor. She performed a phylogenetic of 110 taxa across all sections, and she then tied that back to consistent morphological traits for each section, with a clear identification key. After adding the sections to iNat, I populated them based on the Inacio paper, and on other primary sources that provide unambiguous guidance. Despite that, there are still 48 species not placed in any section at this point.
- Echeandia has similar identification challenges, with 80+ species in the iNat taxonomy. Robert Cruden was the main 20th-century researcher working with this genus, and in 1999 he divided it into two subgenera. Nevertheless, I’m content to leave the iNat genus undivided for now. Cruden’s subgenera seem like a work in progress; unsurprisingly, he wasn’t able to ground them through DNA sequencing, so they’re based on combinations of morphological and phenological characters. These may be valid, but they’re far from clear cut. Subgenus Mscavea generally includes low-altitude species with white flowers that open late morning or afternoon and have narrow inner tepals. But some species in subgenus Echeandia also have white flowers and others also grow at low elevations. And some species in subgenus Mscavea grow at moderately high elevations or have yellow or orange flowers. So the distinctions are messy and unhelpful, and it seems better to wait for future research.
- Polianthes, with 24 species, illustrates another reason not to split a genus. Here about half the species have moth-pollinated, erect, white or pale pink flowers and fall into subgenus Polianthes, and the other half have hummingbird-pollinated, pendulous flowers in shades of red, yellow, orange, pink, or purple and fall into subgenus Bravoa. So, the distinction is really clear. Despite that, adding these subdivisions wouldn’t serve much benefit. Almost all observations on iNat can be identified to species with reasonable confidence, so including the two subgenera would provide little value and add needless complexity.
Overall, I’d say a minimum requirement is to have a valid, published subgeneric framework. But almost as important is that the divisions are useful for iNat identifiers, and that we can expect them to stay fairly stable (i.e. they’re based on strong phylogenetic evidence).