Flag for Taxon: dicots (Class Magnoliopsida)
| Flagger | Content Author | Content | Reason | Flag Created | Resolved by | Resolution |
|---|---|---|---|---|---|---|
| astrobirder | dicots (Class Magnoliopsida) |
Can the APG IV system of spliting this group several ways be followed? |
May. 22, 2018 00:45:10 +0000 | loarie |
see comments |
Comments
That'd be nice; it would require a lot of resources to commit though. We'd be moving a lot of observations.
Fair enough, but if that is the reason not to do it, then why do anything that requires resources?
All of it does not have to be done simultaneously. Different sections can be moved over several days or weeks. After all we are waiting long enough already, spreading it out wont make any difference.
i agree, i can help
How are we doing this?
Subphylum Angiospermae
Class: Basal angiosperms : Amborellales Nymphaeales Austrobaileyales
Class Magnoliids: Canellales Laurales Magnoliales Piperales
Class Chlorantids - Chloranthales
Class Monocots: Acorales Alismatales Arecales Asparagales Commelinales Dioscoreales Liliales Pandanales Petrosaviales Poales Zingiberales
Class Eudicots
. . Subclass: Basal Eudicots - Ceratophyllales Buxales Proteales Ranunculales Trochodendrales Dilleniales Gunnerales
. . Subclass: Rosids:
. . - - Infraclass: Basal Rosids - Saxifragales
. . - - Infraclass: Vitalids - Vitales
. . - - Infraclass: Fabids - Cucurbitales Fabales Fagales Rosales Zygophyllales Celastrales Malpighiales
. . - - Infraclass: Malvids - Brassicales Geraniales Huerteales Malvales Myrtales Picramniales Sapindales
. . Subclass: Asterids:
. . - - Infraclass: Paleaoasterids - Berberidopsidales Caryophyllales Santalales
. . - - Infraclass: Basal Asterids - Cornales Ericales
. . - - Infraclass: Lamiids - Icacinales Metteniusales Garryales Gentianales Boraginales Vahliales Solanales Lamiales
. . - - Infraclass: Campanulids - Apiales Aquifoliales Asterales Bruniales Dipsacales Escalloniales Paracryphiales
We can rename what’s currently labeled as “Dicots” to “Eudicots” and move the orders that are not supposed to be in there out into their correct classes
OK: I can create https://www.inaturalist.org/taxa/872633-Basal-angiosperms but I cannot edit Amborellales Nymphaeales Austrobaileyales to make this their parent.
I suspect that will apply to all of us.
So I have created:
Class Basal Dicots
Class Eudicots
Subclass Basal Eudicots
Subclass Rosids
Subclass Asterids
An issue is if we should keep this totally informal, or go semiformal with
-opsida for the classes
-idae for the subclasses &
So:
Class Nymphaopsida
Class Asteropsida
Subclass Proteidae
Subclass Rosidae
Subclass Asteridae
I think that we should stop there and get buyin from admin and other curators though!
@loarie @bouteloua
hi Tony et al.,
While I'm not opposed to reworking the basal taxonomy of vascular plants, doing so will be very disruptive since it will touch hundreds of tens of millions of observations and identifications so we'll likely have to do it through the server side.
Also, I strongly feel that these kind of major updates to the taxonomy should be done all at once not in parts like this. Leaving around orphaned duplicate/ambiguous classes etc introduces lots of problems.
To that end, I'm doing the following:
1) I made new taxon frameworks for dicots and monocots that don't have taxon curators (ie any curator can continue to edit them)
2) That means the vascular plant taxon framework is now supersceded by 3 downstream taxon frameworks on (a) ferns, (b) dicots, (c) monocots that any curator can edit. But I added myself as a taxon curator for the vascular plant taxon framework meaning only I (or any other taxon curators for that framework we can add) can edit the taxonomy covered by that framework (ie things like adding classes to the angiosperm subphylum etc) - we might want to peel off more taxon frameworks for each of Cycads, Ginkgos, Gnetophytes, Lycophytes, Conifers if having these editable only by the Vascular plant taxon framework taxon curators is a bottleneck
3) I've curated back towards just 2 classes (monocots and dicots) below angiosperms for now (ie swapped Eudicots with Tracheophyta)
Also for what its worth, altering all the IDs associated with dicots is going to be very disruptive since there are over 100k current identifications at this node. I know its not monophyletic (neither is reptiles) but I do think its a useful node for identifications (I personally use it a lot). Also ITIS, GBIF, and Catalog of Life still use Monocots/Dicots as classes. Thats why I'm not personally too excited about prioritizing making the classes on angiosperms monophyletic. But maybe I'm in the minority. Anyway these issues should be discussed (maybe on the new Discourse forum) before beginning the process of curating away from the status quo of having Monocots/Dicots as the 2 classes below Angiosperms
Does this sound good?
yes, sounds good, and i agree, doing it on separate will make more problems than solutions.
I am happy to see it done through the server side.
When can it be done?
I dont think you will find much support for dicots - monocots. Magnolids are a very easy to recognize group too.
So
Class: Basal angiosperms : (3 orders)
Class Magnoliids: (4 orders)
Class Chlorantids - (1 order)
Class Monocots: (already exists)
Class Eudicots: (already exists: all except for 8 above orders carved off).
We further divide the latter to 3 subclasses:
Subclass Basal Eudicots (8 orders)
Subclass Rosids
Subclass Asterids
and Split both the latter subclasses to 4 infraclasses.
Most people will not even notice the difference, but those who are interested in such things will. The only changes are that we are adding 2 levels between class and order, and removing a small number of orders (with few families and species) from Eudicots to three other classes.
At the species, generic and family levels, nothing will change and users will not be affected.
The database indices will have to be reconstructed, but that can be done in the background, or in parallel on another server, or some other way, without necessarily affecting anything, except those who happen to be filtering in the class to order taxon range during the re-index..
I dont the reptile story is quite the same. Almost everybody will initially object if you make birds reptiles, but I dont think too many will object if you make split diapsids into turtles, lizards and snakes, crocodiles, dinosaurs and birds (i.e. carve turtles and crocodiles out of reptiles). I suspect pressure will soon start mounting to align birds and crocodiles as separate from the other clades: it is just a matter of time before you will have to adapt to it.
I suggest you start a thread on the Discourse group and see if there's support for your proposal.
Dicots can't be reused as Eudicots because it includes IDs by people who intended Chlorantids, Magnoliids, or basal angiosperms, so dicots will have to be swapped into Angiosperms which will directly impact over 100k identifications meaning lots of people will notice
I see the reptile story as 100% analogous. Reptiles and Dicots are not monophlyetic, the solution would be to peel off some groups (turtles & crocodiles in the case of reptiles and
Class Magnoliids, Chlorantids, and Basal angiosperms in the case of dicots) to make the remaining clade monophlyetic. In both instances most major Biodiversity databases (ie Catalog of Life and GBIF) still retain these non-monophyletic clades, also there are lots of identifications on these non-monophyletic nodes (reptiles and dicots) and they're useful in helping get observations more finely identified.
I don't feel strongly about either but I'm not personally in favor of prioritizing moving away from dicots or reptiles (I'd much rather get the tips of the taxonomy of both of these groups better sorted). But I'm curious what others think, and whether my instinct that the community will be split is true.
There are only 1,186 observations Id'd as Dicot above rank order. I would be happy to assign them to Chlorantids, Magnoliids, or basal angiosperms, or Eudicots for you.
We could try to go through all of the observations set at Dicot sp. to move them to a lower rank too, making the number of observations that have to be moved to new ranks smaller. On the other hand, the obervations at Dicot sp. would just be moved to Angiosperms, which isn't that much of a difference compared to Dicots.
I spent some time working on the structure of the Vascular Plant Taxon Frameworks https://www.inaturalist.org/taxa/211194/taxonomy_details and
Cycadopsida, Ginkgoopsida, Gnetopsida, Pinopsida are now completely curated
I explained the current status of Dicots in its Taxon Framework Relationship https://www.inaturalist.org/taxa/47124/taxonomy_details linking here
We've still got 30,890 'Relationship unknown' taxa within the Dicot class - I'd personally love to prioritize making some progress towards getting at least some of those families completely curated. But if the community wants to prioritize the basal taxonomy I'm fine with being overruled
Would updating the basal taxonomy not also partition the outstanding tasks into several baskets which can be tackled in smaller groups, rather than just a single dicot-monocot pair?
Surely adding 16 new taxa and aligning their 60 orders is a rather trivial task compared to 30,890 relationships unknown taxa? But could it not be used as an opportunity to bin and assign some of these tasks to volunteers?
i volunteer, just tell me exactly what to do
changes to the Angiosperm lower taxonomy are possible but would have to be rolled out very carefully and strategically. Anyone opposed to tonyrebelo's proposal 'Class: Basal angiosperms : (3 orders) etc...' as what we are aiming for?
no objections? start implementing in four stages?
(have things settled after the CNC?)
@loarie - please check if this is efficient for reindexing the data. Might it perhaps be better to do it in reverse order Infraclasses before classes?
Split Angiosperms to:
Class: Basal angiosperms : (3 orders)
Class Magnoliids: (4 orders)
Class Chlorantids - (1 order)
Class Monocots: (already exists)
Class Eudicots: (already exists: all except for 8 above orders carved off).
Class Eudicots: divide into 3 subclasses:
Subclass Basal Eudicots (8 orders)
Subclass Rosids
Subclass Asterids
Subclass Rosids
Split into 4 infraclasses.
Subclass Asterids
Split into 4 infraclasses.
@tonyrebelo would you be willing to share this Flag in the Forum to cast a wider net? I have no explicit objections, but I'm curious to know what others might think.
I am avoiding the forums - I create too much chaos. Please feel free to take it there.
I'm happy to run it by the forum but looks good to me
also raised here: https://www.inaturalist.org/flags/410810 & https://www.inaturalist.org/flags/410821 - and also related proposal here https://www.inaturalist.org/flags/422171 to add 'spermatophytina' by @carnifex
Spermatophytina is so general as to be meaningless: please dont add it just for lazy identifiers. We need IDs of plants to family level before they are really useful - at least in the more speciose parts of the world. The useful classes are Magnolids (multiple flower parts), Monocots (flower parts in 3s), Eudicots (flower parts in 4s or 5s): it takes quite a bit more effort to get to the subclasses and infraclasses, but when one has got to family level understanding, they are very useful bins for unfamiliar taxa and families.
Interesting conversation. It would really be nice if iNaturalist could be updated to the APG4 classification that we are all using and teaching. I just reminded my students the other day that just monocots and dicots don't exist anymore, as there are basal angiosperms etc before monocots.
This is a tricky one. First, its not 100% agreed upon that we should abandon dicots. Both GBIF, ITIS, and Catalog of Life still include Dicots. They/we also include lots of other known non-monophyletic groups like Reptiles that people don't seem as hostile towards.
But second and more importantly, rearranging this clade would have to be planned out long in advance and would be difficult for us to coordinate because (a) it would impact so many observations, but more importantly (b) the number one computer vision prediction at the moment is Dicots since the model kind of sucks at plant families and I guess the dicot/monocot thing is relatively easy for it to sort out. So changing this would have to coincide with changing our computer vision model which currently takes us many months to train. Otherwise this would probably break alot of things that use computer vision (e.g. Seek, various parts of iNat).
Not to say that this shouldn't, can't be done. Just that every time we think about how to do it seems like a many months, high stakes coordination process for minimal benefit and we end up kicking it down the road
At some stage you are going to have to bite the bullet. Rather now while iNaturalist is still small. Every year the problem will get much bigger. Schedule it for before the next update of the vision module. And get Seek on board for an update then too.
And you are making it a much bigger issue than it is: Monocots stay monocots: it is the dicots that become basals versus multi-part versus 4-5 part: so it is not as big a structural thing as you are making out. In addition the basal are a motley bunch, and not very big.
Basal Angiosperms are 3 orders, 22,374 observations, 78 species (of which 11,373 observations and 32 species are in the genus Nymphaea)
So this is the only "upsetting part of the monocot-dicot dichotomy, and these are quite a distinctive group in being different from all other dicots (and monocots) both in leaf and flower (apart from some Palaeadicots) - so the AI should not have any major hiccups.
https://www.inaturalist.org/observations?place_id=any&taxon_ids=152141,51120,71273
The other dicot groups are straightforward: a simple split into three - those with with the typical 4-5 pattern (Eudicots), and these two groups:
Class Chlorantids - 862 observations of 19 species (only 3 species qualify for AI training). I dont know them at all, but 3 species is not an issue surely.
https://www.inaturalist.org/observations?place_id=any&taxon_id=153069
Class Magnoliids: this is a biggie - but it is firmly in the dicots now and was previously, but it has multiple flower parts as a rule, and may well be easily segregated by the AI as a consequence: in which case this will be really useful. The 126,859 observations (1% of dicots) for 1,424 species (2%) cannot thus be a major issue.
https://www.inaturalist.org/observations?place_id=any&taxon_ids=71057,48808,50898,60244&view=species
((personally it is the Rosid-Asterid dichotomy that perplexes me the most in terms of GISS, would love it if the AI could crack this))
Let us get on with it. A little tweaking of the classification in the future will probably be largely in obscure groups.
I dont think we should look to other sites for guidance on these issues. Let them rather look to us ...
Checking in on this thread to learn more about what would be involved:
Issue 1: To get this to work with iNat functionality we need scientific names can you fill in what those would be in this "Rank Scientific_Name (English Common Name)" format:
Subphylum Angiospermae (Angiosperms)
. Class ??? (Magnoliids)
. Class ??? (Chlorantids)
. Class Liliopsida (Monocots)
. Class ??? (Eudicots)
. . Subclass: ??? (Basal Eudicots)
. . Subclass: ??? (Rosids):
. . . Infraclass: ??? (Basal Rosids)
. . . Infraclass: ??? (Vitalids)
. . . Infraclass: ??? (Fabids)
. . . Infraclass: ??? (Malvids)
. . Subclass: ??? (Asterids):
. . . Infraclass ??? (Paleaoasterids)
. . . Infraclass ??? (Basal Asterids)
. . . Infraclass ??? (Lamiids)
. . . Infraclass ??? (Campanulids)
Issue 2: I agree that it would be fantastic if we could just change the meaning of the current "Class Magnoliopsida (Dicots)" to mean the new "Class ??? (Eudicots)" rather than formally split it into 4 Classes since there are so many existing IDs of Dicots that the doing the latter would require replacing to accomodate a tiny number of taxa peeled off of Dicots. However, that approach will mean that a lot of existing IDs of "Class Magnoliopsida (Dicots)" that were considered "correct" of plants like Magnolias will be interpreted as incorrect disagreements. e.g. if the ID history is
ID 1: Class Magnoliopsida (Dicots)
ID 2: Species Magnolia grandifolia (Southern Magnolia)
ID 3: Species Magnolia grandifolia (Southern Magnolia)
this will go from a current Research grade observation at Magnolia grandifolia to a Needs ID observation at Subphylum Angiospermae with ID1 disagreeing with ID2 and ID3.
Are we ok with this?
Issue 3: There may be other parts of the site that could be impacted by changing the meaning of Dicots. For example the CV model and the Simplified tree on lifelists. I'd need to do an audit of all these and make sure they don't break were Dicots suddenly to change its meaning. I suspect it might make the taxa peeled off from Dicots (e.g. Laurales) behave oddly for up to a year things like the CV model catch up.
Issue 4: The only remaining issue will be a big reindexing job to accomodate the altered ancestries and recalculating things like community IDs which would be triggered by inserting the new nodes (e.g. Infraclass), moving some taxa (e.g. Laurales to new Class ??? (Magnoliids)).
If there's clarity on Issue 1 and Issue 2, I can do an audit on Issue 3 and look into how big of a processing job is involved in Issue 4.
Personally I see the advantage of this change with regards to having a more modern root to the iNat taxonomy. But my concerns remain that its a lot of disruption that won't benefit IDs much (I doubt many people will use these nodes in the ID process since these clades are so phenotypically diverse), and iNat and other informatics sites (like GBIF Catalog of Life etc) will still have lots of non-monophyletic nodes near the base of the tree of life (e.g. Reptiles, Protista, etc.) so it won't solve that issue. I think the big benefit here would be keeping botanists engaged, but that might be naive considering they'd still have Kew POWO taxonomy forced on them which seems to be a deal breaker for many.
These are outside of the formal classification: we are just using the class and subclass and infraclass for convenience because adding many ranks of "clades" is not desirable. Nor do we want to create new names. So we should use the existing names with the corrected rank endings.
Here is a suggestion:
(with class ~opsida, Subclass ~idae & Infraclass ~oidae (this latter is novel - there is no standard infraclass ending for plants), and basal replaced by "Eu~". The insertion of "id" indicates it is derived from clade names, rather than directly from the generic names):
Subphylum Angiospermae (Angiosperms)
Class Magnoliopsida (Magnoliids) [Note: issue here: Magnoliopsida senso lato exists. This is senso stricto. Best not to create a novel name, but it will require a duplicate name on iNat: senso lato inactive (and split into 3 classes) and senso stricto active]
Class Chloranthidopsida (Chloranthids)
Class Liliopsida (Monocots)
Class Eudicotopsida (Eudicots)
Subclass: Eudicotidae (Basal Eudicots)
Subclass: Rosididae (Rosids) [Note Rosidae (subclass) exists but is not used by iNat (except as a synonym to Rosaceae, which is not correct under any interpretation), and would be senso lato as per the Cronquist system]
Infraclass: Eurosidoidae (Basal Rosids)
Infraclass: Vitalidoidae (Vitalids)
Infraclass: Fabidoidae (Fabids)
Infraclass: Malvidoidae (Malvids)
Subclass: Asterididae (Asterids) [Note Asteridae (subclass) exists but is not used by iNat except as a mispelling of Asteriidae - Starfish), and would be senso lato as per the Cronquist system]
Infraclass Paleaoasteridoidae (Paleaoasterids)
Infraclass Euasteroidae (Basal Asterids)
Infraclass Lamiidoidae (Lamiids)
Infraclass Campanulidoidae (Campanulids)
Any further thoughts on this? It's been a while since someone commented here.
While I agree this change would be an improvement, ancestry changes effect every observation in the clade and they're not becoming feasible for nodes at the base of our two biggest clades (plants and insects) without some infrastructure changes.
There's an analogous flag to this about the base of the insect clade https://www.inaturalist.org/flags/422296
I'm going to close this for now and I'll reopen it if/when we can figure out a way to make things more efficient so that changes that impact ancestries of about 1/3 of all obs doesn't create huge problems for the site.
Thanks for your understanding about this - its definitely a problem we're going to have to solve in the next year or two as the site continues to grow
I've also linked to this flag from the deviation for Vascular plant orders and classes (since these aren't in POWO) for reference https://www.inaturalist.org/taxon_framework_relationships/365458
Just now getting caught up on this (currently closed) flag.
regarding names like Class Eudicotopsida (Eudicots)...
... there is almost certainly an existing valid Class name for this clade, and all of the other Class and Subclass taxa in the proposed classification. It doesn't matter that they might have had different circumscriptions in the past, only that the type of the name is included among the descendants in the clade. Likewise, there may already be an existing informal Infraclass and/or Superorder classification from which we can pull applicable names. My main point being, I don't think we need to invent new names (or at least very few) to make this work.
Can we find a compromise? I mean, now (1) we set up subclasses, infraclasses and/or superorders, but (2) we still retain the current Class Magnoliopsida, and ranks like "Chloranthids" could be temporarily kept under it; besides, (3) let viewers see a note (like the note on https://www.inaturalist.org/taxa/207605-Phallus-rubicundus) on the top of the Magnoliopsida page: "Heads up: This taxon is widely accepted but not monophyletic. We suggest you avoid identifying observations as this taxon." Even (4) in the AI suggestions we can use a note or dialog box "This taxon is not monophyletic. (optional words: Would you mind choosing its upper taxon?)". In this way we seem to just create new things rather than delete or transform anything.
Meanwhile iNat is exploding in observations, and making changes like this becomes far more onerous on the system. We should probably push for such changes, rather than wait until it is totally impossible??
@loarie would it be possible to use the "sneakier" methods of editing taxon ranks/names to streamline this process, pushing out the more site-taxing swaps during a time of low use, like immediately prior to Christmas? It would be the best present iNat could give to us taxonomy nerds
not until we find a way to make these taxonomy changes more efficient. We are working today on trying to figure out why even this change https://www.inaturalist.org/taxon_changes/123569 runs out of memory and won't finish which involves many fewer obs than at the base of the plant tree (there are currently 1,190,783 IDs on the node Dicots)
to be clear there are 2 issues here:
1) changing the ancestry of all downstream obs which has to happen anytime the ancestry changes
2) replacing IDs on a node which has to happen if a node is inactivated due to a swap/split
the sneaky thing to do would be to narrow the meaning of dicots without splitting/swapping (which would avoid 2 but still incur 1 assuming some other change in ancestry) but thats still extremely controversal as you experienced with the Jack-in-the-pulplit narrowing-without-split incedent with many users preferring we adhere to our policy. Our current policy is dicots should definitely be split based on the number of IDs involved that will become unintended disagreements. Our system can't currently handle a swap of this scale
@loarie One reason there are still Phalacrocorax auritus observations because some people opted out of both taxon changes and community ID. When someone does the former, their IDs are not automatically changed. And when someone does the latter, other people can't fix the observation's taxon. So if someone did both and they observed Phalacrocax auritus (when iNaturalist called it that), their observation will be Phalacrocax auritus not Nannopterum auritum until they change/withdraw their ID.
When looking for Phalacrocax auritus Research Grade photos, I found only one. The associated observation is from a suspended user, so that's why iNaturalist didn't update the IDs on it. To find this observation, you must use the photos interface, not the observations one.
I agree: that would be most helpful!