On observations with unusually low location accuracy

What is the acceptable minimum location accuracy for a research grade observation? I’ve seen observations where users have created observations with low location accuracy, and/or set the middle of the circle in an imprecise area (IE the organism was observed inside the circle, but not at the center of the circle.)

Whether it is intentional or not is a different discussion, but is the correct protocol to mark “location is not accurate” for said observations?

Is there an acceptable minimum location accuracy? Is an approximately 500km radius too high for research grade?

Thank you. Sorry if this has been posted before, I was unable to find any resources on this specific scenario.

4 Likes

No, if the true location of the observation is within the circle of uncertainty, by definition it is not incorrect.

Not realising that a location consists of multiple data points, such as latitude, longitude, elevation and position uncertainty, is a common user error, but not an error of the data.

BTW if the observation had to be at the centre of the circle of uncertainty, said circle would be meaningless.

4 Likes

What if it is intentional?

Why is this a problem?

My personal interpretation of the “location is not accurate” DQA vote is for situations where the location is clearly not correct, but generally more because the observer uploaded mistaken coordinates. For example a copy of an old analog photo taken in Africa, but which has the coordinates of the photographer’s home in America.

7 Likes

My personal opinion is that polluting data sources with inaccurate data goes against the spirit of citizen science.

Obscured and private geoprivacy settings exist in the case of wishing to protect locations.

2 Likes

If the organism falls within the circle, you should not mark the location as inaccurate.

If the data don’t work for your project, you don’t need to include them. Consider using the “acc_below” search filter to exclude some results. https://www.inaturalist.org/pages/search+urls#location-accuracy

15 Likes

Sometimes a person uploads an observation from years ago where they don’t remember the location. For a couple of my over 7000 observations I have observations with 500 km accuracy. All I know for these observations was it occurred in Ontario (or a minor dip in Quebec) because I haven’t been further than that. With my most recent camera it geotags my images if it gets a lock. If I don’t get a lock I estimate location based on observations where I got a lock.

6 Likes

Accuracy and precision are two different things. The “location is accurate” question is asking if it’s accurate, not if it’s precise - it has to be correct, but doesn’t have to be a small circle.

23 Likes

Note that a custom obscuration (e.g. dropping the pin near the true location while enlarging deliberately the ‘accuracy circle’ accordingly, up to several hundreds of meters/yards) is still much, much better than iNat’s heavy-handed obscuration/private system – in terms of identifiability / interpretability / usability of data (real-life example: to identify some ultra-endemic or geology-dependent cryptic plant taxa).

Should obscured obs/taxa be considered “unworthy of citizen science” or “pollution”? (and as such, barred from reaching RG / entering GBIF?) I fear this would largely invisiblize many endangered taxa…

7 Likes

Doesn’t the standard of “no area size limit” imply that any observation with an unknown location could theoretically be considered “known, and research grade” by simply making the area the size of planet Earth?

Having some sort of maximum area size limit definitely seems more in the spirit of things. Like, if all you know is it’s “somewhere in the Eastern Hemisphere”, does this observation actually have an accurate locality in any meaningful sense?

I think annotators should use their reasonable judgement on this question, and not strictly treat any observation that has a circle as a having an “accurate location”.

4 Likes

Thank you.

If the observation was in fact made inside the mapped circle it is accurate although it might not be precise. If the observation was made outside of the mapped circle, it is inaccurate. Or if it is a very small circle but in the wrong place it might be precise but not accurate.

11 Likes

I know an iNatter who posts records using the phone app and they routinely end up with huge circles around their mapped locations. I know the center point of the circle is the correct location but don’t know why they routinely generate these huge circles when they post a record. I pointed this out to them once and don’t know if they’ve corrected this in their uploading process. It isn’t intentional, just an artifact of whatever process they use.

2 Likes

acc_below_or_unknown is better to use, becouse acc_below filter out observations without accuracy (from phones or photos tagged with gps track).

Maybe we will get this on the website too in future.
That country called Africa - is really not a useful location.

4 Likes

I still disagree with some of this. This doesnt mean i “abuse” DQA. But in an extreme example, you can apparently set the accuracy to nearly what ever you want. I set a dandelion observation of mine to an accuracy of about 25,000km. This is larger than the planet earth meaning the circle is even wrapping over itself. How does that even work or look like?

So for an extreme example like this, are we all saying it could be placed anywhere on planet earth and using location inaccurate is DQA abuse?

While extreme, it highlights the issue of there being no limit, not even a limit to the size of the planet where an observation also could technically be placed anywhere and get RG.

Also tested it with the distance of the earth and the moon. So apparently i have another dandelion observation with an accuracy circle so large it is wrapping over itself possibly hundreds of times.

I dont think there is an actual limit, but pushing it further, setting it to one AU (the distance between the sun and earth). It actually just breaks the editing observation page when you try to save it throwing up a something went wrong.

Edit.
At an absolute minimum i think there should be a hard limit of the size of the accuracy circle that prevents it from wrapping over itself becuase its larger than the planet. But even then the largest circles allowed with that could be placed anywhere and be considered accurate.

5 Likes

The primary issue for me with low location accuracy has been brought up before; private observations exist, which is about equivalent to an accuracy at or exceeding the planet’s size. It’s when the observation is popping up in records that it doesn’t apply to that I get tetchy. That species has exclusively occurred in the far south, where your accuracy circle includes - why are you coming up as this far north, observation? I know exactly where you should be, and it certainly isn’t there.

2 Likes

FWIW: I see no problem with a location set as large as an hemisphere, which might be the largest “accuracy” sensible in a plausible real-world situation: an insect seen onboard a JFK-Singapore flight (… having experienced delays/rerouting, therefore no location deduced from flightplan)

4 Likes

Would you mark such an observation as Casual to keep it out of Research Grade?

Im sure other use cases exist also for hemisphere sized circles, but at a certain point, the larger the circle gets you essentially start saying. Its somewhere on earth, it could be anywhere except Australia. As it gets larger, you begin to question if the accuracy circle stays on the planet, or goes out into space to encompass other objects instead of wrapping over and over on the planet.

But with a policy of, if its in the circle, its accurate. How do you interpret an observation with an accuracy circle large enough the moon is in it if the circle extends into space off the planet.