maybe better for another thread - but we’re close to having what we need to ignore obs with huge accuracies using e.g. https://www.inaturalist.org/observations?acc_above=5000
What threshold should we use? No single value would be perfect in all cases, but the coastal buffer on standard places is around 5 to 10 km. 5km? Only about 3% of verifiable obs have accuracies > 5km
What happens with records where the species is obscured and then unobscured like we have been facing here in Canada. On the records, it seems to show a 20+ km buffer, but it seems that might not be what the user has put in ?
Since atlases use standard admin level 0,1,2 places and these places use true coordinates (not obscured) coordinates to filter observations, issues with obscuration shouldn’t impact atlases (but let me know if you see otherwise).
Remember for all other places that aren’t admin level 0,1,2 standard places (or City Nature Challenge Places) if the place is small relative to the obscuration cell, then the place might not filter observations using their true coordinates due to concerns that people will use small places to locate rare species
I should have been more clear in what I asked. What does the system that reads the buffer ‘think’ the buffer is for this record :
https://www.inaturalist.org/observations/21573673
The text says 27.44 km, but the pin and circle clearly suggest it is not that. This seems to be related to the species being unobscured, then obscured, and then unobscured again in short order.
Maybe we’re not seeing the same thing, or I’m not understanding correctly. 27.44 km seems to match the accuracy circle I’m seeing on https://www.inaturalist.org/observations/21573673
That circle is only about 100 meters in diameter, the whole peninsula that it is on is only 5km long, a 27km circle would extend well inland off Lake Ontario.
OK I see your point, that record still had public_positional_accuracy: 27443 even though positional_accuracy: 122. I re-saved and now public_positional_accuracy: 122
If you can find a way to reproduce please open a bug thread
I dont think I can reproduce it without changing obscuring on species, but it is not just that record, all the ones in Ontario where we went through this open/obscured/open show the same thing:
https://www.inaturalist.org/observations/21604594
The problem isn’t strictly with obscuring, it’s the combination of obscured and large accuracy values. With obscured observations, there’s an accuracy value given, but the circle itself isn’t visible (see https://forum.inaturalist.org/t/obscuring-rectangle-on-obs-with-large-accuracy-circles-may-not-encompass-true-location/703/19 for a suggestion on how to fix that). For observations that are both obscured and with large accuracy, it’s unclear whether it’s appropriate to mark not accurate in the DQA or to let them continue flagging the atlas. It’s possible the accuracy circle does include part of the atlas even if the “true coordinates” aren’t in it. (If that didn’t make sense, I can send a better description privately so as not to “give away” any tips.)
Related, I’m not sure I agree with how you’re using “true coordinates”. For something with a 1000 km accuracy circle, the true coordinates are the center, but that doesn’t guarantee the organism was seen anywhere near there.
At any rate, a 5-10 km cutoff should fix it for both obscured and unobscured observations.
