This is the best algorithm
Well, if you wanted to be completely mathematically rigorous about it, rounding the corners results in actually the smallest possible shape:
But that’s probably a complete waste of effort 
But yeah, as long as it isn’t a problem computationally, it would be best to apply this to all obscured observations. Even observations with very small accuracy values can sometimes be in the wrong box if they happen to fall very close to the edge.
True, but it sounded like maybe there’s a desire to have the rectangle shape consistently mean “obscured”.
Yes, one of the goals @kueda listed was to convey the fact that the observation is obscured, and the rounded corners, while mathematically correct, would be hard to distinguish visually from an accuracy circle.
@kueda any further thoughts? Just saw another observation clearly depicting the wrong location.
i guess i’m having a hard time understanding why this matters. With an uncertainty circle of over 400 km (!) what is it you are using this data for that is spatial in nature? Anything with that low a level of precision shouldn’t be showing on the maps anyway (and i thought it didn’t?). I get that it’s a weird result of the auto obscuring but i kinda don’t get why these observations are being used for anything other than AI training or something anyway.
It’s been said above ^
@charlie I believe iNat needs to take a line then on the largest inaccuracy allowable on a taxon before it becomes useless to all the data partners or the site. Then taxa with too uch inaccuracy, such as the entirety of Yellowstone, do not belong here. If they are useful nontheless to somebody, then conceivably somebody wants to display them. Even if not, I suppose the users might want to display them, because the accuracy might be good enough on the scale of the county or country or region, and so help with the layman’s id, which is what most of the ids are going to be eventually.
I have observations going to a place, which is a rough estimate of where I found them. I could put most of them within 100 yards, but not closer, due to no GPS on camera. I put them much less close than this because observations track human movement and I don’t want to be tracked myself so closely. Should these be used on the site? I move my pins a little too. While the observations have large circles, the circles are much, much smaller than a few levels of classification of the regions of interest where they are located.
For these purposes, an obscuration and no fudging of the accuracy circle might work better, for the consistent randomization it brings. Still there is the limit of whether I know the true true location very well.
And on this discussion, what actually are these observations used for? Who uses them with very fine spatial resolution (and how often does one get that), and who uses them requiring only broad resolution, and who in between?
There was a thread on this, on making the super coarse ones not count as research grade. It’s something i support and it was actually the case for a while but got reversed and I doubt it gets put back. I’m not going to try to explain the reasoning behind it because if i do it will be a sarcastic snarky mess, needless to say it wasn’t the outcome I felt was best. Here’s the thread https://forum.inaturalist.org/t/dont-let-an-observation-attain-research-grade-if-its-location-is-very-imprecise/2072/59 it is locked so you can’t post on it but, you can see the discussion. There is a threshold beyond which it won’t show on the range maps which is good, but the threshold is too high.
please don’t do that, please do create the correct uncertainty circle,because otherwise if you give people permission to use your obscured data points they will not know the data wasn’t precise.
I do! But for like, applied ecology and land management, i won’t be publishing any papers.
i guess some people do though i don’t understand how or why. There’s lots of discussion in that thread.
precise (+/- 2m) = I do, for specimens at arboretum, which are in turn used in studies
imprecise = I do, if I am looking to see what can be found where I am visiting, so I can find and add to my lifelist! I don’t NEED to find it though, it just gives me a heads up on what to look out for.
@kueda any chance for a yes or no on this one? Here’s another case where the user drew a circle around the entire country and the obscuring rectangle now says there are elephants where there are most certainly not elephants: https://www.inaturalist.org/observations/35028234
I think the algorithm from https://forum.inaturalist.org/t/obscuring-rectangle-on-obs-with-large-accuracy-circles-may-not-encompass-true-location/703/20 should work.
I recently saw an observation that had a location accuracy value of 42 km, yet the observer said in the description “Location information obscured because precise data unavailable.”
The result is that while the observer-recorded accuracy circle is 42 km, the obscuring rectangle is 28 km. The observer has in fact reduced the displayed error by obscuring the observation even though they intended for the location to show a larger error.
Just chiming in that this remains a problem. Recent case:
I am surprised this isn’t implemented to date? and now I am wondering if some obscurations rectangles that confounded me while I was IDing are actually completely wrong locations as result of this UI bug.
(as your elephants observation example - https://www.inaturalist.org/observations/35028234)
Location accuracy has been edited since ? Still obscured but a reasonable rectangle now.
I think what happened as this bug report for that elephant is:
user picked a 400km radius circle with no idea where in that park → someone said it is too wide circle → user still cannot (i assume) recall on where exactly they saw in that region → but they used obscure on top of this original 400km radius.
iNat ends up displaying rectangle on top of their 400km using center of that circle for 30km or so obscure rectangle here (As obscure is taking full precedence discounting accuracy circle uncertainty - atleast that’s what I understood and idk if its solved or not), when in fact user did not said that.
aka that 30km rectangle is not what they said at all but they said 400km circle has to be obscured?
I believe location for protected species like elephants are automatically obscured.
So in this case it’s not really an issue at all.
To address the OP, users should not use both large accuracy circles AND obscure the location, as this make no sense at all, if they do it then they shouldn’t complain about a lack of accuracy.
Maybe that could be prevented by the software, a small javascript function with a warning popup like “Reduce your accuracy radius to select obscure location”…
oh yes
not fully, its still an UI-UX issue as this bug report is highlighting. so when I (and often many users) looks readily at obscured observation maps UI as you browse through some set, the meaning such square conveys is feasibility region aka “the true observation is in this rectangle” is the standard assumption of that rectangle for any iNat user.
So with this issue, now every user always has to check Accuracy (and you always have to click details on observation page if not in ID mode just to see “Accuracy” number), because given a random obscured rectangle cell map to you in UI, you cant certainly conclude whether the rectangle you are seeing is original feasible region or actually wrong feasible region as it ended up from this bug and then reconstructing feasible region is not simply draw circle of accuracy value mentally back on to that new point.
well with the elephant example here, even when not done by user, it still causes this bug. ahh got it. I wrongly assumed iNat auto-obscured that elephant observation. seems its not the case when I tested with new observation of endangered species. it is not auto-obscuring (correctly) when the radius is beyond cell size Ah NVM, I havent noticed eye icon (to show autoobscured) and I then opened observation in private tab to see auto-obscure acted automatically as taxon geoprivacy. so this bug as it is is genuine even if not done by users explicitly as large circle+obscure.
actually it is not just simple UI bug as users can subtly miss the true reality, suppose the true elephant is in this circle as drawn below and user forget the exact location but now drawn some certainty circle guaranteeing and encompassing that true location
and now this center of the circle is used for obscuring and we got a nice new rectangle as obscure but say the obscured final coordinate is that red dot there, now u as user see that rectangle and that red dot, if one is eager and say I have accuracy, let me draw 400km around that red dot, you will still miss the elephant’s true point (which observed user correctly said first but you lost that info from this bug) then :(
and so u have to calculate new circle mentally again with 400km+cell size added to it as new radius at that latitude to encompass the true elephant which is not directly apparent to users in first place.
Yes, another, perhaps bigger and centred on the rectangle location, uncertainty 2d shape could be drawn.
And that’s exactly why no one should EVER use a large circle to hide the true location, the “fudge” method is a HACK, totally unscientific! It should be reserved for other use cases, maybe you can’t remember the exact coordinates, or the organism was observed over a large area, but never to voluntarily hide the precise location!
If you want to hide the location use obscure (or private), that’s what this method is designed for!
Again, to stop this nonsense, we need an additional, lower degree of obscure, like 1km or mile, or 1/10th of a lat/lon degree to use the same method used to draw that rectangle! Smaller area if you just want to obscure a house/garden, or for a small organism maybe, and larger (current obscure area) if you feel that 1km/1/10° won’t be enough to protect the place/organism!
you are missing the point here. the elephant observer there cant recall where exactly they saw the elephant in first place, so the correct way is obviously they should definitely use large circle to cover the certainty of their region (such as large park name they can recall there(
