I’m not sure but maybe @loarie or @carrieseltzer has done this kind of comparison?
a relevant blog
https://www.inaturalist.org/blog/76606-thank-you-for-helping-generate-most-gbif-records-for-most-species-since-2020
7 Likes
Interesting! Do most herbarium/museum specimens have gbif records, and are they the bulk of non-inat gbif observations from 2020-2023? I feel like at some point the growth rate of the inat (+ebird/other citizen science) dataset has to be larger than the growth rate for professional deposits for most things, so it should eclipse those eventually.
1 Like
I split this from the original topic.
2 Likes
No, not even remotely close. This recent paper provides a rough estimate that, of the specimens held by 73 of the largest institutions globally (although they are certainly not the 73 largest by any stretch of the imagination), just 16% have been digitised.
I don’t have stats to back this up, but I think the answer would also be no, and that the many citizen science records from other platforms such as eBird would easily outweigh specimen records in that time period, especially given it covers the ‘covid era’
6 Likes
This varies hugely by the specific institution. Wealthy countries have put a fair amount of funding towards digitizing records, as have wealthier institutions. Smaller institutions, those with fewer resources, or those in less resourced countries have much lower rates of digitization. Of course, in many cases, the records held by these institutions might be more valuable than those already digitized. In my opinion (don’t know of any data), digitization also varies quite a bit by taxa. Vertebrate collections often have fewer specimens in a collection than invert collections (probably for purely logistical reasons), so it’s much less work to digitize a vertebrate collection than an invert one in some cases.
A huge proportion of recent GBIF records are from eBird, BUT these really aren’t comparable to actual specimen records in my opinion. They are occurrence data certainly, but require so much less work to process. iNat records are somewhere in between in terms of how much effort they require - they have a type of digital specimen (photo or sound). Actually creating a record with a physical specimen is orders of magnitude more time intensive than either making iNat or eBird records, so there’s no way that there will be as many recent specimen based records in NHCs.
6 Likes
Right, I think the core distinction you are making is: can the ID be revised without intervention by the original observer (other than by atlas) if taxonomic concepts are redefined? If so, specimen voucher (inat, physical specimens). If not, occurrence record (ebird).
1 Like
The herbarium at Oregon State University, by far the largest in the state, has 167,000 vascular plant specimens from Oregon, and about 350,000-400,000 total vascular plants (from everywhere). iNaturalist has about 587,310 observations of vascular plants in Oregon.
6 Likes
Generally speaking, I think it is safe to say that iNat will always have more records of species that can be identified via photos. You have a whole userbase of interested folks looking, observing, photographing, and documenting flora and fauna on a mass scale.
4 Likes
350,000 - 4,000,000 is quite a wide range. Is there a stray zero in there perhaps?
4 Likes
Oops! Corrected now. Thank you for pointing this out.
2 Likes
I used to work at the Botanical Research Institute of Texas, and we had roughly 1 million specimens – when we were digitzing specimens, it was all based on our funding for specific projects. Rather than going A-Z with every specimen, we would have funding for “all vascular plants from X county in Texas.”
Also, on a more somber note, this herbarium was one of the ones that was absorbing a whole slew of ‘orphaned herbaria’ – those collections from universities that could not properly be curated anymore…
7 Likes
I was interested in your question so here is a way to get at the answer for the world and for North America. Citizen science dominates the observations shared from 2020 - 2023 if you look at the occurrences per dataset metric.
2 Likes
This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.