Over the past 5 years, most scientific journals have increased controls and requirements for depositing data/code used in a publication to be banked in an independent 3rd party repository. Dryad, GitHub, and Open Science Framework are some common examples.
I have also considered using iNat to bank verified species observations used in my published studies. See the Notes here for a few examples:
Are there recommendations for best practices in this regard? Linking the DOI is an obvious priority, but are there better ways to tag, group, and upload such observations? Journals like to have a simple DOI or URL for such repos, for example.
I believe that the recommended way to get a DOI for iNat observations is to go through GBIF. If you can set the filters on GBIF to show the observation/s you want, you can download there and it will provide a citeable DOI.
For single observations, I have just used the iNaturalist observation URL and haven’t had any pushback. If you do get this, you could also accession your record as a photo voucher in a natural history collection that accepts those and then use the specimen number. I’ve done this before and still use that method for some publications that require it.
I’m not aware that iNaturalist has a plan for data storage of the sort expected for a repository, and GBIF DOIs for downloads are next to useless because they don’t guarantee the persistence of the actual data (which rather contradicts the point of DOIs). Perhaps you could download the relevant iNat observations from GBIF, add the actual images (I don’t think they are included in the GBIF download) then upload everything to Zenodo. That gets you a DOI with some assurance of that the actual data will persist, including the image. Given that it’s based on a GBIF download the data itself will be in a standard format. Perhaps iNat could offer a service to package a set of observations and upload those to Zenodo, rather like GitHub does for software?
I agree that GBID DOIs are not linked to the dataset in perpetuity (which is a weakness), but I also think that they allow proper citation of collections as well as iNat users which is definitely valuable. It’s easy enough for users to make their GBIF downloaded datasets available for reproducibility on FigShare, etc.
I think this proposed approach to manually upload everything to a repository could work for low numbers of observations but it would be more difficult for large datasets. Part of the issue would be downloading the photos separately (though I think this can be down via the API but not GBIF). It could also be an issue if the licenses for photos are different from those for data. I expect that they’d be the same for most users, but it’s conceivable that some photos might have more restrictive licenses (like all rights reserved) and be unable to be republished.
In the past, I believe that iNat has said that they don’t want to offer easy bulk downloads of images so that iNat isn’t used as a backup service, a goal which might conflict with packaging of observations for downloads.