Given how geoprivacy works by randomizing publicly-displayed coordinates, it is unavoidable that various observations will appear to have a (public) geolocation which doesn’t look accurate at all (e.g. terrestrial organisms in water bodies etc.), and will eventually be deemed “inaccurate” (DQA “location accurate” false) by unaware members.
Is it intended (or even envisioned) by the DQA? Is downvoting this DQA for obscured obs tolerated, and left up to “the community” to eventually revert it?
As long as the true location could be somewhere in the obscured rectangle, that seems like obvious misuse of the DQA.
I would have assumed that by the time someone learns about using the DQA, they also understand the difference in symbols for exact vs. obscured coordinates. Maybe not.
Of course, if it shows an elephant on the Serengeti and the obscured rectangle is in Michigan, we know the location is inaccurate.
This seems not so obvious to a few seasoned users actually … and much to my surprise, there’s little in the guidelines to even characterize “DQAinq obscured obs” as a “misuse”. On the contrary, the example given (“monkey in the sea”) is explicitly not a misuse.
Agreed. The one time things get dicey is if an observation is both obscured and has a huge accuracy bubble. For example, a land animal might have a point dropped in the middle of a huge sea with a big accuracy bubble encompassing the whole seashore around it because the observer knows they saw it somewhere along the shores of that sea but they forget exactly where. If the observation is also obscured, the whole bounding box might be within the sea, making it really look like the location is entered wrong. But in this case the true location is in fact within the accuracy bubble given, but obscuring makes both the dropped point and the accuracy bubble disappear, so it’s pretty much impossible to possess the knowledge to vote correctly on the DQA.
That’s why I specified “on the Serengeti.” An elephant can be captive, the Serengeti cannot, and there is no zoo in Michigan with a convincing replica.
Oh good point. In that case this is another DQA that we really need clearer guidance on.
I would lean toward still downvoting “location is accurate” if the entire visible obscured rectangle is clearly in the wrong place, but be willing to withdraw that vote if the user explains that they actually used a now-invisible huge accuracy circle.
Or maybe we should always ask first, and wait a reasonable amount of time for a response? I’m not sure.
In my experience many users don’t [know how to] ask – they downvote things as they go without letting anyone know, overlooking (or misinterpreting?) the ‘accuracy’ circle or ‘obscured’ status.
Maybe a less ambiguous example could be chosen for the DQA help page (“a giraffe in the middle of Greenland” maybe?): monkeys (and orchids) can (and do) legitimately pop up in marine waters, because of their accuracy circle or obscuration.
It never occurred to me before to think of the monkey in the ocean example as ambiguous, because I took that to mean an actual, un-obscured pinned location of a monkey with it’s entire accuracy circle in the ocean, which is the only thing that actually says, “this monkey was in the ocean.”
If people are using it to mean the randomized coordinates generated by obscuring are “inaccurate” then I agree something needs to be done. For one thing, even if the obscured rectangle is half jungle and half ocean, and the randomized marker drops somewhere in the jungle, there would be an infinitesimally small chance of it having dropped on the true coordinates. So a randomized marker in the ocean half of the rectangle is just as “accurate” as a randomized marker in the jungle half, as long as the monkey was somewhere in the rectangle. Neither means “the monkey was exactly here” they both mean “the monkey was somewhere in this rectangle.”
Or maybe it has nothing to do with where the randomized marker drops, and people are just using it to mean, “I don’t like obscured observations.” There was a thread here on the forum some time back about people using that DQA on observations that had larger accuracy circles than they personally found reasonable. It is a bit confusing that iNat uses the term “accurate” for both. If we had a “precision circle” and a DQA for “location is correct” perhaps it would cut down on that particular misuse.
I think the whole DQA system could be improved if each item came with a hover message: “Use this in cases of X. Do not use in cases of Y or Z.” (Where the latter would cover some common mistakes or outright abuses.)
I haven’t personally seen this particular issue with people downvoting “location is accurate” on obscured observations, but since I rarely look at casual, I guess I would be missing them.
One problem with using words like “accuracy” and “precision” is that it sets up a sort of counter-intuitive situation where a large value is undesirable (whereas “high precision/accuracy” IS generally regarded as desirable).
IMO a better term would be “coordinate uncertainty”. This is what it actually is, and it is more intuitive. ie. it is intuitive that a large number for uncertainty is undesirable.
So many people also don’t realise that precise and accurate are different things.
π=3 is perfectly accurate but very imprecise for example. It’s two orders of magnitude less precise than most usual least precise value used for π: 3.14.
So returning to these observations: perhaps there’s an argument for requiring a given coordinate precision before obscuring is available. That way an observation with a precision of ±200 km cannot be obscured whereas one with a precision of ±20 km can be obscured.
To be clear: accuracy is where the circle is centred and precision is the radius of the circle.
Why not allowing obscuration for very imprecise observations? No point. The very impreciseness essentially already does the vast majority of the job of obscuration.
This is a difficult situation, but is generally rare, thankfully.
I don’t think it’s reasonable to downvote the “Location is Accurate” for obscured observations unless there’s some clear reasoning why it is expected that the true location is not in the bounding box (the same as any observation location really). I don’t think it would be reasonable to downvote simply because the random point happens to be in the ocean when it is clearly an observation on land for instance, as long as the bounding box includes some land where it is reasonable for the organism to occur.
If someone downloads data for obscured locations, they get an accuracy (error circle) value that is equal to the diagonal of the bounding box which is conservative to error (it is the maximum error in the scenario where the true location was one corner of the box and the random point was the diagonally opposite corner). So the accuracy circle for the random point will always contain the true point and the entirety of the box as well.
That said, if there’s an obscured observation obviously taken on land with the whole box in the ocean, downvote away, the same as one would do for an unobscured observation with the whole accuracy circle in the ocean.
Yes, I realize that precision and accuracy are different things, though many don’t. That’s why I suggest that iNat call it “coordinate uncertainty” rather than “coordinate accuracy”.
For the record, I feel that obscuration is probably overused, but I don’t want to get into a debate about that. I generally treat obscured observations as non-existent (I never see them). So as far as I’m concerned, whether or not it is contingent on the accuracy figure assigned to the original observation is largely academic.
But for argument’s sake, I would say that the problem with making obscuration contingent on a specified level of precision/accuracy is that many iNat users don’t really understand it, and probably just use whatever default value is filled in either by their phones, or by iNat (I don’t know all the permutations for how this value is set because I rarely submit observations to iNat). Or they inadvertently set it to either a ridiculously low value, or a ridiculously large value (even though the coordinates are exactly where they made the observation). I see observations that don’t have an accuracy value at all. If the obscuration is contingent on the accuracy/precision value, then anyone who is keen to obscure their observations will simply adjust the accuracy value below the maximum value allowed, and then obscure their observations.
As an alternative to formal obscuration, I often recommend that folks just place their location pin at a non-descript road intersection near the true location (as close as they are comfortable with), and then adjust the accuracy/precision/uncertainty radius so that it encompasses the true location. This gives the observer control over just how precise they want to be about specifying the location of their observation, and prevents “accidental” disclosure of a location that the observer feels the need to conceal. This alternative approach has been discussed in the forum at length, and many participants are dead set against it, but I feel that as long as the “fake” location is within a kilometer or two of the true location, and the uncertainty radius encompasses the true location, it’s perfectly acceptable. Personally, I don’t record the precise location of every organism I report, so it’s not unusual for my own observations to have coordinate uncertainty of several kilometers. YMMV.
They’re of course better defined as a measure of precision than accuracy, and I know the terminology has come up in a few discussions over the years. The pedant in me would be happy if they changed the term to “Precision”.
I skimmed the thread but did not read every post thoroughly so I apologize if I missed this, but are we talking about downvoting everything that is obscured. Seems harsh to down vote all orchids when they are only obscured due to collector concerns. Same with other endangered species.
This observation does in fact have a correct location. When you click on it and zoom in on my location, you’ll see an obscuration box in the middle of the Baltic Sea, which looks wrong. But look how big the accuracy bubble is:
I did in fact observe this plant within that circle. But if you can’t see the circle due to the location being obscured, you just see a box in the middle of the Sea, and it looks to all the world that the location must be wrong. So what are you to do?
As @cthawley said this is luckily a rare situation, but it’s the reason I personally avoid voting on the location DQA for obscured observations. How do I really know that the location circle is wrong if I can’t see it?
Obviously there are rarely cases where anyone would do this (I just did it as an illustration and I’ll delete it once this thread closes), but it definitely happens.
everything that is obscured: downvoting looks like a misuse of the ‘Location accurate’ DQA
some obscured observations: downvoting might be appropriate in some cases (cf. Solution to this thread) although it would be hard to ascertain
non-obscured observations: downvoting is appropriate if pinpoint+circle are clearly absurd (monkey-in-the-ocean example)
As cautious observers, we should routinely monitor our ‘casual’ bin. Users (including seasoned ones) poorly aware of the ‘accuracy circle’ and ‘randomized pinpoint’ concepts can (and will) cause stealth casualization, by downvoting this ‘Location accurate’ DQA.
An identifier may be able to see the real coordinates of an obscured observation because they are curator of a project which the observer has joined and granted permission to. If they flag an observation for inaccurate coordinates, it may appear to those who can’t see the true coordinates that this is unjustified. I can recall specific scenarios where this could have happened.
… and much to my surprise, there’s little in the guidelines to even characterize “DQAinq obscured obs” as a “misuse”. On the contrary, the example given (“monkey in the sea”) is explicitly not a misuse.
