Edited my response to get a broader overview
Edit
Also looked into the first Senecio example. It got a explicit disagreement from species to complex by a very experienced identifier. I would not have much hope that it will be identified to species.
Is a personal lack of āhopeā reason enough to prevent future identifiers from confirming a specific ID?
edit: btw, those Senecios are already and will be identified+RGed to species!.. as soon as a second knowledgeable identifier arrives. The fact that some identifier disagrees with a previous identifier (or with the observer) certainly does not mean that all hope is lost to identify the obs - especially where only one species exists.
It would not prevent identifying if they are at research grade.
Use the filter I provided and start doing it.
It largely hides them from most identifiersā sight, by removing them from the āNeeds IDā pool while they a) still need ID, and more importantly b) still could be identified down to finer levels with very little effort. This latter point contradicts the opening premiss that complexes are used āwhere similar and related species are barely distinguishableā - this is a faulty generalization when it comes to iNaturalist, given the meaning and usage of āComplexā on this site.
Making the (too) few, overburdened identifiers have to constantly chase through more supraspecific ranks (in addition to casuals, genus-level obs DQAd manually etc.), to eventually fish and re-ID among the many, many obs having been automatically and wrongly deemed unidentifiableā¦ is of little help. The fact that wrongly RGd complex-level obs would eventually be locatable and salvageable by a select few dedicated identifiers (who master complex sets of filters) does not really help. Flooding GBIF with data at genus-level and with an invalid (orange), misleading name attached - even if temporarily, until some iNat identifier finds time and energy to properly locate and re-identify thousands of RGād Complex obs - is perhaps not the best outcome either.
Weāre going in increasingly vague circles I fear. I exposed many counter-examples and a few arguments, letās hope (!) it was for the best.
edit: as a compromise, in order to leave skilled identifiers a chance to arrive and do their magic, why not implement a mandatory delay - to automatically RG any obs that has been stuck at infrageneric-to-supraspecific level for, letās say, one or two years? I think Iāve already seen a feature request like that.
There are lots of species complexes on iNat where some species within the complex are readily identifiable at least some of the time (in certain regions or certain sexes/life stages etc.) ā so it would not be a suitable solution for all or necessarily even most species complexes.
Even for the example you mentioned of the complex Achillea millefolium, not all of the species included within this complex on iNat are impossible to distinguish from field photos (A. pannonica, A. roseoalba). Likewise, the complex Lotus corniculatus contains Lotus tenuis and Lotus alpinus but not Lotus pedunculatus, even though all three are arguably similarly easy or difficult to distinguish from Lotus corniculatus.
This proposal would also require not insignificant changes to the way the taxon and status displayed for an observation is handled on iNat and specifically the way this interacts with the community taxon. Right now if someone enters an ID that is finer than the current ID on an RG observation, one of two things happens:
-
if the observation is RG at species and someone adds a subspecies, the species continues to be the observation taxon (the taxon shown at the top of the page and in searching etc.) until there is a second subspecies ID.
-
if the observation is RG at a level above species because someone has checked the box āID cannot be improvedā and someone adds a species ID, the observation taxon will be the species and the observation will be displayed (and shared with GBIF!) as āRGā under the species ID, even though according to the community taxon (the 2/3+ consensus) the RG should only apply to the genus ID.
It seems to me that either one of these behaviors would not be desirable in a case where there are thousands of observations being automatically made RG, some of which will almost certainly be able to be refined further.
It may be that changing the way that the community vs. observation taxon is displayed and searchable on iNat would be useful in any case ā but it would be essential in the context of this feature request.
1 Like
I realy like this idea. During the whole discussion some have argued that they would not like the somehow reduced visibility of RG observations, thinking this might slow down the IDing process. On the other hand, including all infrageneric taxa would avoid some other questions like: How explain average iNat user the difference between a complex an a subgeneric taxon? What should happen if a complex would be raised to a subgeneric taxon and vice versa?
This adresses an important point. I was thinkig about this too. I agree that RG should not be reached by automaticallly change DQA to cannot be improved. First to avoid the side effects. And second it would complicate the DQA procces would need to be to be explained to observers.
As result of our discussion I would alter my initial proposal as follows:
Elevate infrageneric-to-supraspecific observations automatically to research grade following the rules that would apply for species if they stuck at of those levels a year after the had been uploaded.
.
Addiontionals thoughts: I think a year is good enough to give IDers time who a specific season to do indoor work time. On the other hand itās an easy to explain reasoning. I would start to count the time with uploading date so as to avoid complicated extra rules.
.
2 Likes
In my experience my plant observations are typically identified the next winter after I post them. Insects might get to family or tribe in that time, but for many insects and most fungi itās common to wait 2-3 years before someone comes and identifies them to genus (longer to get to species, if ever). Lichens are even worse, essentially never being identified beyond my initial ID (usually at class level). I donāt think time-based filters make sense for this.
3 Likes
Obviously it depends on the region. At least for fungi a takes a few days in my area to bring most of them to species or at least complex even when they start at something like āfungusā.
However as proposed observations at genus level and above would not be affected but only observations that surpass the 2/3 quorum at a level that is narrower than genus.
I assume I now would now need some input from a staff member. As I would see it, the discussion about wether or how it would be a good idea calmed down with the change of my initial proposal as cited below. I think this might be a good point to discuss the idea inside the iNaturalist team. Or didnāt I understand the process?
As result of our discussion I would alter my initial proposal as follows:
Elevate infrageneric-to-supraspecific observations automatically to research grade following the rules that would apply for species if they stuck at of those levels a year after the had been uploaded.
.
Addiontionals thoughts: I think a year is good enough to give IDers time who a specific season to do indoor work time. On the other hand itās an easy to explain reasoning. I would start to count the time with uploading date so as to avoid complicated extra rules.
Agreed. Iāve been sorting observations of moths from oldest-to-newest the past few weeks, and Iāve found there are tens of thousands of perfectly identifiable ones sitting unidentified from 8+ years ago, even in a well-covered area like the USA. There are simply too few people identifying them to have any sort of expectation that IDs will probably be provided within any given time frame.
7 Likes
Those ones sitting unidentified for so long, are they sitting at complex/section/infragen-supraspec rank? or are they just at order/family/genus rank and therefore not involved?
1 Like
I believe this would require some sort of process to run in the background, checking probably thousands of observations or more every day, and potentially slowing things down. I donāt think it would be worth it, to be honest.
4 Likes
Uhm. I think it would be worth it. I can imagine there is a solution that is not too expensive. E.g there are similar processes when observations are processed to go to GBIF and others.
Currently 1,236 pages of āNeeds IDā complex/section moths for the USA alone, 700 pages older than 2 years. Scrolling through the 2+ year old ones, there are lots that can be identified to species. Most are cases where two people added disagreeing IDs of related species, the CID got kicked to ācomplexā or āsectionā, and no further IDs have been added since then.
Either we need to click through these pages and manually click ācanāt be improvedā on the ones that truly canāt go to species, or weād need to manually click through all the āauto-research-gradedā ones and manually add species IDs to the ones that can be taken further. Thereās just no reasonable automated method to differentiate between these two categories. Personally, Iād rather hold them at Needs ID until someone RGās them with a ācanāt be improvedā vote than send them all to RG and then try to pull back all the ones that really did need ID attention.
5 Likes
115 Likes.
I still think itās worth to evaluate the amount of work that is needed.
1 Like
Looks like 115 likes refers to total number of likes on everyoneās replies to this topic. There are 9 votes for the proposal.
5 Likes
Ah. I am not sure how to read that number. Thx
1 Like
This one is problematic. At least one of our local experts idās as Complex Centaurea jacea specimen that are broader, hybrids with other species in sect. Phrygia (or Jacea). In other publications the complex C. jacea agg. is understood more broadly.
1 Like
I would not support this, because in the moths (Lepidoptera) we have species complexes where its species can easily be IDāed for caterpillars or by dissection. The complex is merely for the adult moths which were just photographed.
However, can I suggest to avoid any species within a complex to be used for the AI model? It occurs to me that several moth species will automatically be identified to species level, but they should (without dissection or caterpillar info) always be at complex level. It looks like complexes are not used for training the AI model. This causes a lot of work to actually modify all IDs of common moth complexes, or what I observe, IDāers are simply neglecting them.
Examples:
(1) complex Mesapamea secalis: the AI model uses the species Mesapamea secalis, but should stick to the complex all the time. Species indistinguishable on photos. (currently 11,600 observations)
(2) complex Oligia strigilis: the AI model seems to use the species Oligia latruncula and Oligia strigilis, which may or may not be correct. I simply have doubts and would prefer the AI to suggest the complex, not the species. (currently 11,300 observations)
(3) complex Acronicta psi: the model seems to use Acronicta tridens and Acronicta psi, but should stick to the complex all the time. Species indistinguishable on photos.(currently 9,000 observations)
(4) complex Noctua janthina: the model seems to use Noctua janthe and Noctua janthina, which mostly cannot be decided without the underwings. I would prefer the AI to suggest the complex, not the species. (currently 9,200 observations)
Apparently, the complexes are not used on the list when using the ācompareā button, as soon as there is one species IDāed within that complex. This may be a bit more difficult to solve, but it also leads many users to be overconfident about IDāing the species.
3 Likes
It is possible for a complex to be presented in the CV results, but itās very rare. For example on the current website model it suggests the Narceus americanus complex for any millipedes that fall under that complex. However itās a constant chore to prevent N. americanus from getting added back into training, and as soon as it does it starts suggesting just the species again.
3 Likes