There is endless talk on the OSM-Carto issue tracker and here too. I researched a bit and came to the conclusion, that natural tags in our arsenal are not at all orthogonal – in the sense: At all times, only one applies.
The area pictured below as of today is mapped as natural=bare_rock:
In the EU Copernicus Corine dataset the area is classified Sparsely vegetated area. Where I am at home, this type covers 13% of the area of the state. It is not particularly rare
Upon closer inspection, some of the cushions look like this:
It is a bit like a fractal: The picture from afar shows rocks, gravel, grasses, shrubs. The picture close up shows herbaceous plants (grasses, flowers), woody plants (Rhododendron here), gravel. All of these have tags in openstreetmap natural=* vocabulary.
I think natural=fell tries to be orthogonal to bare_rock or scree or even grassland or heath, while it isn’t.
Perhaps an attribute to indicate sparsely_vegetated missing?
Nature mixes landcover types as it pleases, with no regard to mappers I would try to follow clear border lines as much as possible and tag each area by what it mostly is so that a user in that area will be able to orient themselves using the map data (i.e. “according to the map I should follow the path through a grassy area, and that direction looks more grassy than this one, so I should look for a path in the grassy direction”). It is indeed a kind of fractal, just like you could endlessly discuss by how many nodes a curving path or the shore of a lake should be described. I think we should be guided by what is practical for the map user. GPS has an accuracy of about 5 m, so we should not map more detailed than that for sure.
I am fine with bare_rock for this fell there. But maybe we can do better. As this is not the least about rendering, here a mock-up (not true to ground, but showing an idea, real cartographers certainly can beat that):
I think is essential to keep feature size large. We are talking about vast spaces. There is a lot of fell mapped, it is not all bogus. What is mapped fell in my area not converted to something that renders for years now. The other signatures just look wrong. So rather left blank in the Standard View.
I do not think any such differential tagging based on local observations alone, albeit reinforcing a tour description from some book will ever pass muster with the OSM-Carto maintainers.
Certainly, the grass looks greener on the other side, always. Sorry no ads free link found. Taking away fell from people mapping that looks like a hard sell as the Joneses can keep their scree, bare_rock, heath &c. There must be something in for all concerned.
GPS has an accuracy of about 5 m, so we should not map more detailed than that for sure.
what do you mean by this? Should a building 1x2m be mapped the same as one 6x7m? Should we not map paths shorter than 5m? Should we not map any detail smaller than 5m? I don’t think gps accuracy should influence what we map and at which precision.
Seemsike we have density tags in use for most vegetation layers. wood:density for trees and shrubbery:density for shrub layer.
Missing would be the density tag for the herb layer
So maybe we need a herb:density
Of course not, what I wrote is in connection to the preceding sentence about paths and lake shores. We should map as accurate as possible: building dimensions can be measured much more accurately, so a house with bay windows can be mapped with those windows showing because their location and size can be measured with an accuracy of centimetres. But in this post we are discussing how to map landuse in mountainous areas, so there I think it would not be appropriate to map a lake shore around every stone at that shore, if the position of those stones can only be determined with approx. 5 m accuracy.
so there I think it would not be appropriate to map a lake shore around every stone at that shore, if the position of those stones can only be determined with approx. 5 m accuracy.
generally I agree (that we must not exaggerate the shape complexity of lakes), although it should be noted that there are different kinds of accuracy involved, for example if a road is perfectly straight, we can map it with just 2 nodes, or if a building is 7x12m we can map it with these dimensions and rectangular shape, although the position could be off by 5 or 10m, and it might be slightly rotated. More than the absolute positional accuracy of nodes (for which we typically don’t have the tools to obtain very high precision), we care about relative accuracy (things relating to other things in the map, e.g. the side of the road where something is, shape of things, etc.)
I see, the forum is mostly for making jokes that make nobody laugh or for fundamental discussions about geometric precision.
Meanwhile I learned of https://wiki.openstreetmap.org/wiki/Proposal:Stone%3D* - It failed on something, this topic is not concerned with: Here it is rocks with some vegetation - there it is vegetation with some rocks.
Accidentally, natural=fell can mean both.
PS: All the OSM bare_rock and scree and fell in this picture below is CORINE sparsely vegetated. Probably, because Corine requires 25ha minimum feature size, while in OSM one can spend hours to map a single square metre at 2cm resolution? Top post pictures stemming from left summit.
Going by your pictures I would say that the vegetation isn’t dense enough to be natural=scrub or natural=fell. Your second picture looks like alpine tundra (natural=tundra) or natural=scree, depending on whether the sparse vegetation is due to altitude or not. That said, the presence of some vegetation does not preclude the use of natural=bare_rock. So perhaps the answer is to use bare_rock for the areas with minimal vegetation (eg. your first picture) and either scree or tundra for the slightly more vegetated areas around the bare rock.
Not sure if it helps or hinders, but I am having the same issue with coastal dunes(i’m less experienced). By nature sparse vedgeitation, causes coastal dunes to grow, so natural=heath (or =grassland) is appropriate although it’s ~30% sand, adding landform=dune_system seems to work. Whilst landform=* should be paired with Natural=landform, I can’t see it breaking any code. But it feels really clunky. (Natural=dune is also possible, but not rendered, and feels simplistic). Correct me if I’m wrong.
Nice find with the dunes. Where my aunt used to live, there was the beach, then came the dunes (sparsley vegetated, more than half sand) then came the macchia (woods).
I would try to follow clear border lines as much as possible and tag each area by what it mostly is so that a user in that area will be able to orient themselves using the map data (i.e. “according to the map I should follow the path through a grassy area, and that direction looks more grassy than this one, so I should look for a path in the grassy direction”). It is indeed a kind of fractal, just like you could endlessly discuss by how many nodes a curving path or the shore of a lake should be described. I think we should be guided by what is practical for the map user. GPS has an accuracy of about 5 m, so we should not map more detailed than that for sure.
