Geoprivacy, Obscuring, and Auto Obscure Discussion

this is a place to discuss general policy issues and questions regarding geoprivacy, obscuring observations, and auto-obscuring. If you have specific feature requests related to this please post them as feature requests. This is more a place for general discussion.

Please do not post loopholes or weaknesses in the obscuring mechanism in this thread or elsewhere in the forum. For obvious reasons, if you find something like that we don’t want it visible to anyone else. Instead, it’s best to email

Broadly, I made this thread because i think there are a few large scale issues and questions that come up. Feel free to add your own

Firstly, the display of obscured observations can be problematic. I know there are programming and mapping limitations, but the randomly placed point consistently causes confusion. As more observations are added, the obscured observations also totally cover the map. See below, in Los Angeles, for example. (circles obscured)

Secondly, once you zoom out a bit you can’t tell which observations are obscured, which makes range maps and such confusing.

I personally would be in favor of having the obscured observations display like Ebird does when zoomed out (but the non obscured ones stay the same). Ideally the obscured ones would also be in a different (non sticky) layer that could be turned on and off. I don’t know how hard that is to do with the current map engine though. See below again.


Thirdly, it would be nice to tighten up what is auto obscured. Anything with collection and harassment risk should be obscured. However, there is also the odd case of common plant species like white oak being obscured when their range edge clims the edge of an administrative unit. It makes us harder to look at the distribution of white oak in terms of climate change, invasives, etc, if it is obscured on the edge of the range due to an arbitrary overlap with a human created boundary. Of course, if the species has collection or harassment risk, it should be obscured. But without describing what we are doing, it can seem arbitrary. And it isn’t clear who should be contacted if there’s an obscuring issue. I’ve been trying to track down someone in Quebec to ask to unobscure these common tree species, but it isn’t even clear who obscured them in the first place and i don’t think one curator should unilaterally undo it anyway. We need a policy.

I realize this is a hard topic because people have very strong opinions ranging from ‘obscure nothing ever’ to ‘obscure everything always’. I don’t think either of those are a good option, and I think iNaturalist does a good job of walking the balance. I just think we also need a ‘tighter’ way to deal with this.



Reply #1 - I agree it is problematic that any single curator can change a status, I would propose a change so that it requires 2 different curators to make a change.

To me the optimal way of doing it (without knowing how easy this is to program) would be to utilize the taxa change framework and make the change into a formally entered taxa change, one that has a rule that the drafter and the committer have to be different people.


Reply #2 is more a discussion point. to me it remains an extremely significant issue that the site seems unwilling or unable to add national or local red lists or similar frameworks outside North America (I apologize, I’m not sure if it is done for New Zealand).

Why locally imperiled or at risk species in other nations do not have the same protections in terms of obscuring as is done in Canada and the US is a massive issue to me. Adding global frameworks is a poor substitute, there are many species which are globally secure which in turn have significant risks in more local geographies.

My understanding, which may be completely wrong is that there are indexing concerns with adding large numbers of conservation states, but surely then at least something can be implemented that documents the file formats required to load, and times uploading them to a time when an unrelated indexing task is required.


There is a third option (just for completeness).

  1. Randomize obscured localities
  2. Grid represent obscured localities.

The third option is to just autoobscure it by just lopping off the “insignificant” digits. So obscuring to 100km, have coordinates truncated to 0 decimal places (e.g. 34 S , 19 E), and
to 10km: = 1 decimal place (e.g. 34.1 S 19.4 E)
1km = 2 decimal places (e.g. 34.13 S 19.39 E)
100m = 3 decimal places (e.g. 34.134 S 19.394 E)
What this will do is superimpose all the obscured localities onto a single point in the NE, SW, SE, NW corner (depending on which quadrant hemisphere you are in). So hidden localities will display as a grid over the landscape - but diffuse enough as to be non-invasive on the display.
This also has the advantage that by using the number of significant figures one can exclude obscured data from a location unit.
The problem is that viewing/selecting/opening such observations will be difficult, but since they are randomized and thus irrelevant to the display anyway, does this matter?

But I Like the gridded cells option most.
But can I make a request that observations that have a Locality Resolution (aka Accuracy, aka positional_accuracy) higher than critical levels also be obscured. So if a locality resolution is 100km, then the data should display just like hidden data. But I realize that some people (esp. those who do not understand the concept and may have recorded their resolution incorrectly or not at all) may find this troublesome, but conceptually it is the same thing.


This could be very helpful. (Off topic, but I wish it was easier to see location accuracy in the identify tool.)

I would also like it if there were more than one “size” of obscured grid, to deal with the different needs. In many cases I think the current obscured grid is too large (or unnecessary).

I want to use a small grid to hide my field active equipment locations (camera traps, etc.).
I want to use a large grid for species threatened by collection (poaching, etc.).


I second that! Often I have to open an observation outside of the ID tool, only to discover the bad locality is a resolution issue (esp. with cellphones just switched on that dont have a good locality yet).

Must this have a new topic, all will this be extracted from here?

1 Like

I’ll make a new feature request for this:

1 Like

That could work! And would result in less ‘clogging’ of maps.But wouldn’t help with the issue of a point existing where it shouldn’t. But… maybe that’s a mapping software limitation if grids are harder to do.

Maybe I am not understanding something, but would this not reduce the randomization of obscured records from 25km to 100m in many cases? That seems a steep price for cleaner maps.

1 Like

Huh. Yeah didn’t realize that. Is there no rounding that results in something close to the current obscuring?

1 Like

2 additional concerns under the law of unintended consequences

-I fear it would lead to a significant increase in flags for inaccurate locations along the lines of ‘the map shows this on water, clearly it is wrong’

  • more concerning, if everything with 101m or above uncertainty is mapped to a common point of 1km2, I fear a lot of those points will end up on private property with a good risk of people following them onto private property to look for things.

very true, but the private property issue exists with the current system too, it’s less concentrated but affects more landowners (maybe all landowners in some areas?)


If people willingly post stuff from their property and show where it is, that is one thing, making it look like observations are there when they are not is another.

1 Like

sorry if i was unclear, but that is my point too. the current obscured points also show up in the ‘wrong’ place that is also often private property so rounding them would cluster them, but either way they aren’t marked precisely and the map icon implies they are, at least a little.

1 Like

Related thread:

1 Like

It depends on what levels of obscuration are required. We explored this in southern Africa and decided that 100m, 1km, 10km and 100km are suitable levels for most situations, even though historically we used quarter-degree obscuration (i.e. 1:50 000 survey maps, or about 25km) for herbarium data.
Having intermediate levels is not rocket science. Instead of merely lopping off decimals, one can have a system that filters to intermediate levels and rounds off (e.g. instead of 10km and 100km, one could have 30km and 50km and 66km - i.e. x | x.0 becoming x | x.66 | x.5 | x.3 | x.0).

The points about locations now being in the sea or on specific land parcels, applies equally to whatever obscuration system you apply, so is not a game-changer here. And in both cases they are displayed as circles instead of pins. However, in the case of set points, they are predictable and avoidable (i…e one can create a mask and hide the points - if one is grabbing screen data for example). Ditto: the unintended consequences of people marking inaccurate localities, just changes the details, not the proportions or causes, in any predictably biased way)


I would actually consider a totally different approach with one big caveat*

  • for an obscured record, on the observation page itself, don’t show a point map at all, show an administrative map with the appropriate jurisdiction coloured in. If it was seen in Harris county, Texas, show a map that highlights Harris county with text something like ‘the exact location of this record is unavailable for public viewing, either by observer choice or due to the sensitive status of the species. It was seen in Harris County, TX’
  • on the species page turn the checklist places view into a heat map - thats not the right term but colour gradient based on the number of sightings, and make the point map layer only open records
  • make the default view on the explore page to not show obscured point records, and add the toggle to turn them on that has been asked for so many times
  • the one caveat is how this would work with ocean observations where of course there are no administrative boundaries to map to. Ideas on that are welcome.

I dont like the idea of adminstrative boundaries: which level - magisterial, provincial, national. Some magisterial districts are huge, and it would be equivalent to several degrees of obsfuscation. Other areas are very finely divided - too finely to properly hide data. Using such a feature does not make sense. One wants control over the level of obscuration: not to have useful distributional data in poorly populated areas because the units are too big, but to have it too obvious in areas with small units is not acceptable. A grid of some standard size is by far a better approach.


I’m sure this has been discussed before, but why do observations where the location information is completely private go into the “Needs ID” category rather than “Casual”? Unless I am missing something, it seems like from the perspective of anyone trying to help with identification, there is no real difference between “Location Private” and “Location not specified.” Also, the Data Quality Assessment doesn’t allow users to downvote “Location is specified” even if months have gone by without the original poster responding to a query about the general location of the observation.

I guess maybe there’s some benefit to allowing these observations to get to Research Grade if the species is sufficiently distinctive? I’ve just noticed that there seem to be a disproportionate number of “Location Private” observations that stay in the “Needs ID” bin for a long time even though some of them might be identifiable with very basic location info.


The location should have absolutely nothing to do with the identification.
On another thread, I pointed out that we need three separate issues dealt with:

  1. Identification - how sure are we of the name given
  2. Data quality - how good is the data (time, date, locality, picture/sound quality, etc.)
  3. Planted/captive vs wild

Please dont confuse things. Location private or not specified is a data quality issue, and should have no effect at all on the ID quality-score-status (although it might if species or subspecies require locality to make the ID, but purists never do this)…