Consider this observation of a crayfish burrow chimney. It’s identifiable to the family Cambaridae. It has been marked “No, it’s as good is it can be” for the “Based on the evidence, can the Community Taxon be improved?” question. This seems to make the observation Casual. I don’t understand why this should be. It’s perfectly good evidence of the family. I’d think it would be useful to include in the CV model for identifying the family. Does this mean it is not included?
There was a recent change to how that works (I am wary of clicking that)
I think it has to be Genus to make it RG ? Correction : Subfamily.
Staff decided over a decade ago that family would be the cutoff for RG/sending to GBIF:
Anyone getting data directly from iNat is still welcome to use family-level data, it’s just not getting sent to GBIF.
This is incorrect, Casual observations can be used to train the CV, see here for more details.
Anything subfamily and finer can go to RG.
7 Likes
It’s useful at higher taxon levels than family where there is very little hope of moving downward because it removes the observation from the Needs ID queue. I can see this being a challenge for the family level though.
Yeah, there are multiple different species in this area, many of which are in different subfamilies or complexes. Unless someone can pin a chimney construction style to a specific subfamily / complex / genus, I can’t see any way to move it past the shared taxon.
1 Like
Interesting, there’s a different staff explanation here but I guess that one is older:
This is an effective solution to that particular problem, but I think it causes other issues because the “Casual” label has pretty strong semantic implications on iNat as indicating that the observation is useless or trash data. For example there are very strong objections to making observations with accuracy uncertainty in the thousands of kilometres be Casual because there’s a chance they some might have value. Personally this feels inconsistent, so I think if that’s going to be the case then observations stuck at family level (or even higher) should also be able to be RG too since they could also theoretically contain valuable information, and RG status also removes them from the Identify pool…
For what it’s worth, a family-level observation is barely ever going to be used in CV training because the CV only trains on leaf taxa. If there are any genera or species below that family which have enough observations for the CV to train on them, then it will completely ignore any higher ranks. Feature request to change that here: Allow some non-leaf taxa to be added to the CV model
2 Likes
What are leaf taxa?
The terminal nodes in a taxonomic tree. That means different things in different contexts though. If you go to “Your Observations” and click on the Species tab, it shows leaf taxa that you’ve observed. If you have observations of a genus and none of any of its species then the genus will show as the leaf taxon in the table there. (The help page here talks about it in this context)
In the context of CV training for iNat, it’s the lowest rank taxon in the tree (but no lower than species) that has enough observations to add the taxon to the training pool. If one or two species in a genus have enough observations, they’re the leaves. If the genus has enough observations but none of the species do then the genus is the leaf. If nothing below family level has enough then the family might be the leaf.
3 Likes
Thank you!
1 Like