Talk:Corine Land Cover

From OpenStreetMap Wiki
Jump to navigation Jump to search


Corine sample

Take a look to the rough visualization of Corine (attached Corine sample image). Some data (light green bold borders) is existing OSM data for comparison, all other is Corine. I have following general ideas based on this:

  • importing all these polygons would be too much data. The polygons should be merged to larger areas and classes, also tagged accordingly. E.g. I see no value of separation of wood=coniferous and wood=deciduous, it just would generate a lot of extra polygons. Just one landuse=forest would be fine
  • do not import agricultural areas at all. At least in my country these would not be rendered on most street maps in any way, just white areas. Everybody knows that what is not forest, lake or artificial should be some kind of agricultural area. Maybe it is different in some other countries with mountains etc. Maybe some classes (wineyards) would be valuable. This would also save a lot of data.
  • Some other polygons, like classes 111 and 112 should be merged to single areas. It is already suggested to have same tag.

--Jaakl 11:14, 29 May 2009 (UTC)

I think it's better to have "all" the data, renderers can then omit farmlands or whatever it is that is the locals consider "default landscape". There will be massive amounts of data anyway when the map is "finished" so some polygons more won't likely be too much, although it's quite possible to drop the coniferous/deciduous... By the way, how many polygons and points are there (for a country?) if they're not combined? Alv 11:37, 29 May 2009 (UTC)
(I'd too like to see the wood types imported, thought just that those doing the import might not want to. Great if they do. Alv 12:31, 29 May 2009 (UTC))
I'm in favor of tagging differently coniferous and deciduous.
France is 270 000 different polygons, 550 having more than 2000 nodes, 30 having more than 20000 nodes. See details on the french page (numbers for each polygons in brackets) Sletuffe 12:23, 29 May 2009 (UTC)
A small country like Estonia has 35 291 polygons, I have not node statistics. Area of Estonia is according to [Wolfram Alpha] about 11 times smaller than France; and whole EU is about 10x France, so I would expect 3 million polygons in total.

Hum, we have a different approach in France. We don't want to limit the import to the smallest list of common interest from people involved in the discussion (which is anyway a small subset of all contributors). Many contributors would like to see only forest because is it impressive to see the big change on the map (especially in countries like France where forest is extensive) and you really feel that you show something visible from the ground. But we are not importing for the renderers. And farm landuse is not a less valuable information than forest. We also don't want to import stupidly all polygones 1:1 from CLC to OSM but we study each CLC classe and check if it has a value for OSM or not. Fro instance, we mostly reject all classes that are mixing too many different things because the information is too slight. We feel that some categories are just created to fill the gaps. Another point is that we want to preserve as much as possible a link with the original data because we know that every 5 or 6 years a new version of the dataset is published and each time, it is more accurate (minimum change detected is now 5 hectares for CLC2006). So speculate that in the future, the data will tend to be always more accurate and we already think about the future and try to make the job easier for the next updates. About forest, we will keep the "coniferous/deciduous" differences because we see no reason to remove this information . But we will try to simplify by re-using the same nodes for adjacent polygones. And we know that the amount of data will be important but this is nothing compared to the valuable information it is bringing into OSM. -- Pieren 12:21, 29 May 2009 (UTC)

Why destroy perfectly good data by simplifying? Import all the polygons, and preserve all data by including the original CORINE classification using some sort of "corine:class" key/value tag. As was said before, renderers can choose what to display or not! --Petz 12:55, 29 May 2009 (UTC)

The current is using the following CLC:id CLC:code CLC:year in order to keep all the information that we are getting from Corine. The forest north of my town is a mixed of "native" forest (oaks, etc..) and a mix of coniferous that were introduced about 30 - 40 years to help the local industry and to help reconstitute the forest since pines grow faster than oaks. I care pretty strongly about the different areas as some will have some mushrooms and the other not. --Melaskia 00:44, 31 May 2009 (UTC)

Combination with existing areas and later updates

Another question: how do you plan to combine information with existing areas, mainly forests and urban areas? Generally existing manual data is probably more accurate, and there will be a lot of conflicts (overlappings). I planned initially to skip CLC imports for areas which are well covered with existing data, but thinking of "import everything and enable later updates" strategy this is not so simple. Especially thinking about later CLC update options. In any case there will be update conflicts later: say area boundaries will be modified in OSM and will be generally more accurate than CLC later update (which has lower accuracy/scale). But also later update may come with new re-classifications and improvements for the same area. Manual review of all the conflicts then? So two related questions: 1) how to solve conflicts now 2) how to solve when new CLC version will be available? --Jaakl 07:01, 2 June 2009 (UTC)

All polygons overlapping existing areas, all polygons in rejected classes (too fuzzy classes), would be rejected, but stored in a way that is still not fully fixed for a future use. We are talking of a layer on a server to show conflicts, and a plugin-layer in JOSM to handle those polygons. This system, if it prove to be good now, would also be good in the future. FrViPofm 20:23, 3 June 2009 (UTC)

Bare rock, loose rock or bedrock?

The point about bare rock, I do not think natural=rock is the right way to tag it (see discussion on Talk:Proposed features/rock. A better solution IMO is to tag it as natural=bedrock if that what it is or natural=scree if it is loose rocks. --Skippern 17:15, 21 July 2009 (UTC)

Broad-level IS NOT Decidous

Broad-level Corine class includes come evergreen species, such as Querqus Ilex, it can not be equivalent to Decidous. See CLC documentation: