This problem is only getting worse. When I try to ID something in the SW US currently labelled as “Sauria” the map shows ALL the lizards in that area. So, each time I want to see where that observation came from I need to unselect the map option that shows these just to see ANY of the map. A check box in the settings (probably under content and display) that allows users to add or not those layers of the map would be less intrusive than it is now.
1 Like
There is an option that has not been suggested here, but which I think is the most logical and precise way of doing it, and which gets around the creation of false localities (which iNat does and I think is unethical). .
It assumes that obscuration is related to data precision (or location error, what is called location accuracy on iNat [even though the value increases as accuracy decreases]).
Simply, for coordinates:
0 decimal points ~110km error
1 decimal point ~11km error
2 decimal points ~1km error
3 decimal points ~100m error
4 decimal points ~10m error
5 decimal points ~1m error
greater than 5 decimal points - nonsense values
So deprecated locality data due to obscuration can simply be rounded off to the appropriate decimal points. This is the mathematical or scientific way of resolving the issue.
The decision now becomes, should obscured data on iNat be rounded off to 0 or 1 decimal points?
This simply means that the points for any cell are now relocated to the NW point of the map (in the southern old world hemisphere - I presume this varies in the four hemispheres), and the huge clutter on the maps, is now resolved to 1 point every 100km or 10km. These are all neatly arranged in a grid, and it is easy to see which grids have obscured data and which to not on busy maps, although this is not so easy on sparse data maps.
It also means that the downloads of data are scientifically accurate, and we wont have accidental use of the “false” data in publication (both scientific and popular) maps. And data can be shown below the maps without “wrong” coordinates.
In the good old days when iNaturalist had to “pad out” the maps to make them look busy, the current approach was a reasonable (if confusing) compromise.
But now iNaturalist has sufficient data to remove all the clutter and rationalize it, as well as presenting the data more “accurately” based on the obscuration.
Whether these data are displayed on maps as grids or circle-points is arbitrary in my opinion: personally I prefer to have a single graphic for a map feature (i.e. dont mix points and grids).
One outstanding issue, then, is what to do with observations that have a provided location accuracy of over 10km - should these data not also have their finer accuracy values truncated to 0 or 1 decimal points, and displayed as part of the “grid”?
And what about observations without a location accuracy? Are these really trustworthy? ((please do not hijack this topic around this issue: please start another topic if this excites you)).
PS: personally, if I had the choice, for single species maps (not mutliple taxa data) I would skip the pegs and display the localities as circles based on their location errors: smaller (and darker) circles for precise localities and larger and more transparent circles for less precise data (with missing locality accuracy defaulting to 10km). But I understand that this would not appeal to many users.
3 Likes
Maybe I’m missing something, but this system seems like it would still introduce datapoints that are just as “false” as obscuration (I don’t really agree with the usage of this term) - it’s just that all the “false” datapoints would be in one location for any given area.
I also think that the display would be less ideal. At high zoom (on very specific localities), you might see zero points for the species, making it appear as if it is not in an area (if you don’t happen to include the corner point) or users might think it is only in one specific location (the corner point). At “medium” zoom levels, currently, the spread of obscured IDs allows you to get a relative idea of how many observations of a species are in a given area. With this proposed system, this wouldn’t be the case - you would see one point, whether there were 1 or 1000 observations. At very zoomed out levels, I suppose it would be quite similar to the current system.
1 Like
probably it was not discussed because no one wanted to “hijack” the topic to go on some tangent about changing the fundamental obscuration logic.
if you want to represent obscured points visually as a single arbitrary point within any given 0.2 deg lat x 0.2 deg long obscuration box, it seems like a way to “reduce clutter” on the map, but there likely many reasons to avoid doing this, some of which cthawley touched upon. for me, the most obvious reason is that you don’t want whoever is located near such an arbitrary point to all of a sudden get all sorts of random observations showing up at their location in the system. (you can think of the current implementation as prioritizing privacy over possible clutter on the map.)
this doesn’t seem to solve the problem of “clutter” and probably would make the problem worse, besides introducing all sorts of new problems…
i actually agree with this post. True it wouldn’t remove ‘false’ data points but i dislike how they are scattered all over the map. Looking at the map of my observations for instance it looks like i covered a whole rectangle with observations when i really just took 50 in one part of it.
But an easier solution would just be to make it possible to turn on and off obscured points on the map and make them different colors when you zoom out. Right now the circles only show to differentiate from the pins when you are way zoomed in. With the grid view, you get a lot of middle ground where the observations create these ugly confusing squares that would be nice to either not see or just have as a different color in the background. Once you’re zoomed way out to continent level it doesn’t matter of course.
3 Likes