Whoa, whoa, whoa…you mean top-tier! 
(Just joking around, I understand your meaning).
Dear Mehmet @salix-babylonica
I understand very well what you mean, because I am interested on flora of Algeria, and I also regularly take a look on flora of Syria, Iraq or Libya.
My experience is that the situation could change very quickly if one first user (like you on Turkey, Arya in Iraq, or Naser in Libya) shows the way.
In Algeria, friends had been organised a City Nature Challenge few years ago at Constantine, and now the Constantine area is the most known in Algeria ! they also, like myself, encourage new users one by one, both from the amateur communisty and the academic one, and the situation is coming better and better for more and more regions of the country.
Thanks to this incresing actitivy, the flora knowledge is becoming better and better, and I choose to be a curator of the taxonomy few years ago especially for helping that process.
So I think you can keep hope, but for sure you have to be active and motivate all your friends and colleagues to do the same, explaining them what are the (numerous) adventages and the (few) disadvantages of using iNat, both for themselmves and for the biodiversity…
Errol.
I think - @tiwane has said he will do / has done outreach via Zoom - if invited - say to biologists at the university?
Think a lot of this is cultural. iNaturalist simply appeals more to Americans. Obviously you’ll have a number of people from every country that enjoy it, but I find more average Americans enjoy things like this than average non-Americans.
(A bit shocked to realise that California alone has over 39 million people)
South Africa as a whole - about 63 million.
It’s very easy to not realize and keep in mind how large and dense the United States is.
Traveling across it is very much like traveling through many contiguous countries! Language, customs, and laws differ as much as populations … and then there’s all the diversity and abundances of the non-human life!
I think the same could be said for California alone! (-:
Not to mention factors like how obvious the species is (pretty sure if a bright, large flower and a small, unobtrusive white flower are equally common, you’ll get far more observations of the former), how easy it is to identify (particularly if it needs flowers and only flowers for one week in the year), how easy it is to get an identifiable photo (small things will suffer, for example - and good photos of narrow flower stems can be hard to come by), etc.
iNat records indicate presence, but say little about either absence or relative abundance.
I don’t know what these observations were of, but it’s worth bearing in mind that it can be as simple as what species you observe, and sometimes whether the CV recognises it. I’m in Australia and have seen plant observations sometimes identified within seconds, and other times taking years. Going through old Needs ID, I’ve found observations stuck at high levels for 5-10 years that I’ve been able to ID to species on sight, and others identified to species that apparently never got agreed with (okay, or they were agreed with by someone who’s since left) - so it’s not just the area.
Having said that, if you can encourage more people to start identifying, that would be great! :-)
Agree to some extent. Sometimes the “suggestions” are North american species that are suggested without any hesitation. But generally I think it really depends on type of organism and country. Also time of the year as peolpe spend less time in front of a computer in summer than in winter so ID might be delayed a bit. So far I have recieved many good comments and IDs from Australia and othe countries visited.
If the observation has a location when the ID suggested are made, by default iNat should only show suggestions where the species is “expected nearby”. In the past, though, this restriction wasn’t there.
It’s amusing when it gives me a bunch of foreign species that are completely and obviously incorrect but if I just subtract the species and look at local stuff in that genus I can find the exact one immediately. Usually when the local stuff isn’t observed nearly enough to be in the machine view but its very close to the foreign ones, is when its all exotic suggestions. Occasionally I need to step down a few more taxa levels, or machine view just has no idea, but often enough if the photo’s decent it gives a good ballpark.
I’m sometimes asked, and always willing, to try identifying a fly in one of the families that I generally only identify in Canada or the US. I can’t do that unless I can find free online publications with the necessary information. Even then, it could easily take me an hour or more to carefully plough through and compare with the unknown.
Finding more identifiers and observers is a chicken-and-egg problem. People like to know what it is that they’ve found. If no-one ever responds because there are no ID resources, will new observers stay long enough to become identifiers themselves?
Having more resources scanned and uploaded could really help for currently underserved areas. Checklists are handy, but without a key, and if necessary, the original descriptions (hopefully with photos or diagrams), it’s just too difficult or time-consuming to identify in many places on an occasional basis. Likewise, if active researchers could make their articles free to use, that would benefit us. A journal that has exactly what one needs, but is behind a paywall, is next to useless for amateurs like most of us.
E.g., speaking parochially, The Canadian Entomologist and The American Entomologist are mostly inaccessible to non-members of Ent. Soc. Can. and Ent. Soc. Am., respectively. This is very frustrating, as they both go back to the 1860’s, under one name or another. At least the ESC has an ‘Entomology Enthusiast’ membership for $68/yr, bless them, which is affordable for me. ESA has nothing similar, and their regular annual membership dues are a whopping $170 (~Cdn$ 250). Surely they and others could open up their older articles, perhaps those that are at least ten or twenty years old? How often does anyone pay for access to their vintage material, anyway?
I’m still a little confused by this. Earlier this week while identifying spiders in South Korea I came across an observation posted two months ago from the iPhone app where the Computer Vision had recommended a North American species to the user.
Checking the species overview page, that was the only record for South Korea (now gone since I disagreed), and the six other observations for Asia are all from India and Bangladesh. Is that a normal distance for the “expected nearby” to be offering suggestions? I assume there’s something I’m missing, because it doesn’t seem accurate to label a suggestion as ‘nearby’ when it’s from halfway across the world’s largest continent.
(I also noticed that those six observations from the Indian subcontinent are the only ones from outside the Americas, made using the Android app. I recall reading about one of the apps not including the CV icon when users selected one its suggestions, which has me curious if those might be ‘unmarked’ CV suggestions.)
Alternatively, do you think this would be a case where the observer made an observation with no location through the iPhone app, received and agreed with a CV suggestion for the North American species, then later went back and edited the observation to include a location?
Sometimes the CV will make suggestions that are not expected nearby but just visually similar.
You can select whether you want it to show only suggestions that are expected nearby or all possible matches and sometimes people will use the latter. The CV symbol next to the ID is the same regardless.
If the CV is struggling to find any good matches for the photo, it will give some low-confidence suggestions which generally seem to disregard geography.
I don’t find the presence or absence of the CV symbol to be a reliable indicator of whether the ID was inspired by the CV or not. It is possible to use a CV selection without the symbol appearing if you reclick on the selected ID (at least on the website – I don’t know exactly how this works in the app).
The CV also does seem to have a few odd “expected nearby” calculations. There is a N. American species complex that had a single erroneous RG observation in Europe which I corrected at least a year ago, but it still seems to pop up occasionally for European observations with the note “expected nearby” (confirmed by checking what the CV suggested for me on observations where users had selected this ID)
We need a higher barrier for Seen / Expected Nearby than
one obs
with one ID
(which is perhaps wrong).
At least
one at RG
or a few at Needs ID.
@spiphany explained it well, I suspect it was a case of low confidence for anything expected nearby so it just provides a list of visually similar suggestions regardless of location. If you want to share the observation here or with me privately I’d be happy to take a look.
Which one is that? One thing the geomodel is use species seen nearby as part of its prediction, so if a mix of similar species are seen in both locations, that could be an explanation.
I think you’re referring to the way “seen nearby” worked, but that hasn’t been part of iNaturalist since the geomodel came out in September of 2023. It’s just “expected nearby” now.
This one: https://forum.inaturalist.org/t/why-getting-palaearctic-species-suggestions-for-nearctic-observations/48165/15
I am not seeing it as “expected nearby” today on the recent observation that popped up with this ID, but I am quite certain there was one I corrected last week where it did show this. Maybe the most recent CV update finally fixed it. There are still some stray observations for the complex in S. America and SE Asia which I suspect are also wrong but I have not corrected them because I only have species lists for Europe.
(Frustratingly, this is one of the cases where the criteria for what is included in the CV training paradoxically lead to worse suggestions. For most of the observations where it has been suggesting the complex L. gemmatum it at least has the subgenus – Dialictus – correct, whereas once the out-of-range species complex is taken out it suggests a different genus in the same family (Halictus subgenus Seladonia) or something completely different, like Chelostoma. Because Dialictus is one of the more difficult bee groups they are rarely ID’d to species, so the CV is unable to correctly suggest this for the thousands of observations in the subgenus in Europe)
Honestly, I think that has just made me more confused. My understanding of the dialogue has been –
- kstenman posted that the CV suggests North American species when it shouldn’t.
- You posted that, if there is a location, the CV should only suggest species “expected nearby”.
- I posted about my confusion, with an example.
- spiphany posted that sometimes the CV makes suggestions that are not expected nearby but look visually similar.
- You posted that spiphany explained the issue well and added that the CV might list something from far away if it looks visually similar.
I’m having a hard time processing your first and second comments together because it feels like the first is saying that North American species (with a location at the time the CV makes its suggestion) shouldn’t be suggested for other continents while the second seems to be saying that it’s expected for the CV to suggest North American species for other continents if they’re visually similar and it doesn’t have a local species in its data set.
The explanation from spiphany is how I thought the CV operated (leading to “mistakes” of North American species being recommended over local lookalikes due to only the North American species being in the CV data set) but your initial comment in response to kstenman gave me the impression that shouldn’t be the case. I imagine I’m overthinking the whole thing but, as someone who does most of their identifying for observations from outside North America & Europe, this is an issue that impacts me enough where I would like to make sure I understand the process correctly.
Here is the observation that I had referenced in my previous post:
https://www.inaturalist.org/observations/263567436
I had a somewhat similar issue come up recently with Homidia socia, which is a North American species of springtail that the CV is recommending for observations in South Korea and China. After a quick check online, I came across a research paper that listed 46 Homidia species recorded in China, which makes me wonder how many of those CV-suggested IDs may actually be a local species that the CV doesn’t know to suggest.
(Of course, it may be an introduced species in China with each observer knowing enough to distinguish H. socia from local species; I will admit to not asking each user in China if they could support their ID. I did ask a user in South Korea who chose the CV-suggested ID about it and they mentioned they they had just moved to South Korea and didn’t know why they had decided to ID at the species level.)
There are three possible reasons why the CV might be suggesting species that are out of range.
By default, the CV prioritizes suggestions that it thinks are possible in the region based on its geomodel. This will be indicated by a note next to the suggestion (“expected nearby”). However, under certain circumstances it may make suggestions that are marked as “visually similar” but not as “expected nearby”. This happens if 1) the CV has a low level of confidence and cannot find any “expected nearby” suggestions; this should normally be indicated by some text at the top about “we’re not sure but here are some suggestions” or 2) the user changes the suggestion setting to include all suggestions and not just "expected nearby; there is a link at the bottom of the list of suggestions allowing one to do this. My point was that just because someone has used the CV to make an out-of-range ID does not necessarily mean that the CV actually thinks it is a good suggestion for that location – the symbol only indicates whether the CV was used, not whether it was a high confidence or “expected nearby” suggestion.
There is also scenario 3: Occasionally, the CV makes suggestions that it labels as “expected nearby” but are in fact wildly out of range. From what I understand, the geomodel calculates expected range based on a combination of existing observations of that species, certain geographic data such as elevation, and co-occurrence with other species. But this is not perfect and sometimes it seems that it draws wrong conclusions about what is plausible. This seems to be what is happening in the example you linked to, and the case of the North American bee I mentioned above that has been persistently been suggested as “expected nearby” in Europe. I don’t know exactly what causes such miscalculations, though it seems more likely to happen if there are a scattering of wrongly ID’d observations on multiple continents that nobody has cleaned up.
https://www.inaturalist.org/blog/99727-using-the-geomodel-to-highlight-unusual-observations
Since I am on the Cape Peninsula - too small for the blocks - the discussion in the comments is useful. I am using that - geomodel anomaly - thoughtfully - for CNC IDs.