Do Private observations impact observation's utility for research?

I was under the impression that making locations private or obscured was a good way to protect sensitive locality data from the general public, while still making the observation useful for research. I assumed that iNat’s research partners like GBIF could still access this data. But in a comment on another user’s observation someone mentioned that this is not the case, and that private observations don’t have any utility to GBIF, which apparently does not have access to hidden locality data. He appeared to be well-informed, claiming to have worked with GBIF. So what’s up? Was the guy just wrong, or if he was right why are private observations eligible for Research Grade?

2 Likes

Private observations are exported to GBIF but without locality information. e.g. https://www.gbif.org/occurrence/3070523560

There is not a protocol by which researchers can access locality information short of contacting the observer and asking.

As the location information is one of the most important pieces of information associated with an observation, making observations private or obscured definitely decreases their value to research.

17 Likes

Additionally, I think obscured observations get exported with the randomized location. Depending on the research, this could make a big difference. I think GBIF basically counts as the public here.
https://help.inaturalist.org/en/support/solutions/articles/151000169938-what-is-geoprivacy-what-does-it-mean-for-an-observation-to-be-obscured-

Research grade is an iNat designation that may or may not line up with specific research projects. As stated, the observations are still exported but their usefulness is significantly lower.

6 Likes

A large amount of private observations cant even get IDed without the location info to begin with.

13 Likes

Absolutely, private observations have a reduced value for research. That isn’t to say that some observations shouldn’t be private, but there’s definitely a trade-off. Personally, if there’s an observation I feel is so sensitive that I would want to make it private, I generally don’t even post it on iNat.

The question and related ones (about IDing private observations) have come up on the forum before and have good info, examples:
https://forum.inaturalist.org/t/if-location-is-obscured-is-the-observation-still-useful/46522
https://forum.inaturalist.org/t/does-obscuring-your-location-affect-the-chances-of-getting-research-grade-ids/59899

I can confirm that private observation locations are not exported to GBIF - otherwise it would pretty much defeat the purpose as anyone can make a GBIF account and download the info. Users can choose to trust other users or projects with true location info, so it is possible to share that location info for private observations via iNat. As such, there’s some potential utility there, but this only applies to a very small proportion of iNat users.

6 Likes

I thought the same, but seems like they do, see Kevin’s link above

3 Likes

There’s a big difference between Private and Obscure, in that Obscured shows the locations general area, and Private shows no location at all. Having working a little bit with GBIF data, having an Obscured location hasn’t been a big problem for me, but Private locations are more problematic, and I might not use them, depending on what I’m looking at.

For instance, one analysis looked at phenology by latitude. Obscured locations still had roughly the correct latitude, but Private locations wouldn’t work at all.

12 Likes

I often obscure observations since I do a lot of them at my place of residence. Seems like it simply greatly increases the location’s encompassing area but keeps the general location. It may be less useful for research that wants to use smaller area units, say.

2 Likes

For my house I just set a point a few km away, with about several km uncertainty radius such that my house is in the accuracy circle. This way the accuracy is much greater than if I were to obscure and hopefully the observations will be more useful.

2 Likes

If private is to protect your home address?
I made a pinned location.
So far, none of the random dots have actually landed here.
https://www.inaturalist.org/projects/elephant-s-eye-on-false-bay

I do mostly obscure for orchids or unusual bulbs.

2 Likes

if “private” allows you to feel comfortable sharing more, i’d say it benefits research. especially local research where you can privately tell an actual human where the “private” observation was made. trusted local organizations can sometimes be more impactful (especially at the local level) than data analysts looking at global data at a global level.

as local organizations become more familiar with observation based technology, it seems likely that they’ll come up with ancillary databases that store location information for private gbif observations.

The observations themselves are definitely exported, but not the locations, which was the claim I made:

At the link, GBIF notes:
“There is no location information published for this record”

Though interestingly, I noticed that it does give a coordinate uncertainty value:
Coordinate uncertainty in metres 30377

That looks like it could be an obscured uncertainty value, but of course could also be the actual one as well.

@kevinfaccenda out of curiosity, is that value the uncertainty for that observation when you entered it or is it obscured?

1 Like

Any data point aptly recorded and archived somewhere (be it a notebook) is usable by, and potentially useful to, research – be it ‘RG’ or not on some social platform, exported or not into some gov database, geolocated or not. Discoverability of data is certainly better (but by how much, if cluttered?) once concentrated in a few large open repositories; however this is still no guarantee of validity, or quality, or traceability, or usability. It is up to researchers to gather data wherever it may be (museums, agencies, photo/social websites…), then assess it properly (… if ethics matter) according to their own expertise and peculiar goals (not taking any ‘community vote’ and ‘location’ at face value).

I wouldn’t pressure iNat users to “make data available on [other platform]” (by e.g. enforcing RG at any cost) or “prepare data to make it useful to researchers” (hard to foresee what will be of use).

1 Like

This was shocking for me. I was under the impression that obscured observations simply obscured the view of the location for the iNaturalist public - that the most accurate metadata would be accessible in the research dataset. This is bizarre to me if it is true as it can really muck up analyses.

2 Likes

Yes, its mainly for privacy reasons. I have made an addition to my bio that states I can disclose a more precise and accurate location to researchers should they be interested in using my such observations for their work.

1 Like

So obscured location just means “mega-large accuracy area”? Does the size depend on the size of the accuracy circle you set?

Maybe, maybe not. It depends on the parameters of the research. We offer the information, the researcher decides what to use.

1 Like

The research dataset on GBIF is publicly accessible - showing true locations there would defeat the purpose of obscuration. GBIF does showed the obscured locations and the corresponding uncertainty. It’s the responsibility of researchers to use data appropriately (though I have definitely seen researchers use this data incorrectly, but that’s their mistake).

There are other ways that researchers do have access to the true locations of observations with taxon or personal geoprivacy. Major details of how iNat handles geoprivacy are here:
https://help.inaturalist.org/en/support/solutions/articles/151000169938-what-is-geoprivacy-what-does-it-mean-for-an-observation-to-be-obscured-

6 Likes

In order to use records, researchers must first be able to find them.

The likelihood that they find records will be greatly increased if these records and the data associated with them is labelled in some standardized manner and made available for sources that collect data of this sort.

Researchers, like everyone else, have limited time. If they can get large quantities of records that have already been screened to ensure they include certain basic elements and use a standardized format, it is going to be more efficient for them to use records from these sources than go searching for small quantities of data scattered across dozens of other digital or non-digital sources that may not be publicly accessible.

Certain choices – such as a license that does not allow for sharing with GBIF, or choosing “private” for the location – will reduce the findability of one’s observations. These choices may also reduce the number of research questions for which one’s data is useful.

(There may also be other non-trivial pragmatic implications – if an observation is marked as private on iNat, it will not show up in searches for taxon specialists who focus on a particular region, even if the observation includes a text note about the location. If one does not include any information about the location, as I see happen quite often in observations by users who do not understand that “private” hides all location information, most specialists will be reluctant to ID all but the most distinctive organisms. Since many observers rely on specialists to help identify what they found, in both of these cases marking the location as private on iNat may mean that a rare find might never even be recognized as such.)

Noting the ramifications of certain choices allows users to make an informed decision about what they want to do. It is not the same thing as demanding that they behave a certain way.

4 Likes

It seems wild to me that the random location is exported. I understand the use case for the randomized location for the sake of visualization on inat maps, but the randomized location is objectively worse for research purposes than the 0.2 degree rectangle that actually contains the location of the observation.

2 Likes