A series of observations on the border of Ukraine and Slovakia are unaccounted for by both countries, despite falling within the iNat borders designated for Ukraine

Platform: Website

URLs (aka web addresses) of any relevant observations or pages:

Description of problem:
I have a series of observations on the border between Ukraine and Slovakia that are not displayed in the search results for either country. I was expecting that both imprecise iNat polygons are simply not covering this area. However, these observations are actually within the polygon designated for Ukraine on iNat (place_id=8860), 1 km deep (but in reality it’s right on the border), so it’s something else.

Interesting. I can replicate this and verify that it is happening.

The observation location states it’s in Ukraine, but when it is searched for via the place with user and species filters on it does not show up.

As it’s not a new observation it should not be a result of lag in the site updating its information.

1 Like

I have found this to occur when the precision of the observation location meant it could be in either place.


did you already try forcing a reindex on the observation by, say, updating one of the Data Quality Assessment flags?

Yes, the accuracy circle clearly extends across the border for the observation, so this is most likely the issue.

this shouldn’t matter for “standard” places.

you’re right. it actually looks like the system’s boundaries for Ukraine & Slovakia really don’t include the observation. in other words, if anything can be fixed, it would have to be those boundaries.

here’s a map showing the observation within Transcarpathia, UA (the level 2 administrative region within level 1 Ukraine):

here’s another map showing the boundary of Ukraine here (note how it fails to include the spot where the observation would be):

here’s another map showing the boundary of Slovakia here (note how it also fails to include the spot where the observation would be, and note how it complements the boundary for Transcarpathia but not for Ukraine):

the place polygon boundaries that you see in the Explore page on the website are simplified. so you can’t always rely on those as providing a true representation of the boundaries stored in the system. the map tiles (as show in the maps above) provide a better representation of the true place boundaries stored in the system.


Thank you for explanation! I’ll simply move observations several meters to the south and will increase accuracy range, already tested that it works, observations starting to appear in the search right after moving, even without delay. But it’s a shame that precise matching polygons are not used even for countries on iNat. There is probably a lot of such “lost” iNat observations on the national level around the world.

1 Like

yes, unfortunately, the boundaries in iNat can be a little odd in places. i guess the best we can do for now is to wait and hope that one day the boundaries will be updated to address all the different problems that have been pointed out since the last update (especially North America).

1 Like

Is this an iNat issue or something with the source of the polygons for the standard places? I was under the impression that these were derived from another source. I think @jwidness is familiar with some of this.

It’s as pisum said – the shapes came from GADM a while ago, and iNat simplified them on import. iNat then simplified them again, so what you see on Explore isn’t actually what determines whether an observation is in a place or not. Standard places haven’t been updated or corrected in a long time because it would be an extremely computationally expensive process.

“Precise” is relative. iNat’s shapes are far more precise than, e.g. this map:

It’s about striking a balance between file size/computation time, and fine-scale validity (i.e. the more correct it is at a fine scale, the larger the file and longer the computation time), and while I’m sure there are quite a few misplaced observations, they represent only a small fraction of iNat’s total number of observations. (To be clear, I agree that misplaced observations can be extremely frustrating, especially in geopolitically sensitive locations.)


the thing that’s surprising to me is that there are gaps between the neighboring polygons, and that parent and child polygons have different common boundaries. even if you simplify the boundaries, i would have thought that that process would have been done in a way that preserved common boundaries across relatives and neighbors.


As far as I understand it, the processing iNat originally did was supposed to create seamless boundaries. I don’t know where or why it went wrong :woman_shrugging: