[Imports] CLC meadow cleanup

[Ilpo Järvinen]

On Tue, 25 Mar 2014, Martin Koppenhoefer wrote:

> 
> 2014-03-24 10:56 GMT+01:00 Pieren <pieren3 at gmail.com>:
>       Altrough the CLC program is european,
>       it's done by many different operators in a long run. The quality
>       varies from region to region.
> 
> yes, but generally the quality is good :)
> The problem is the scale, CORINE is in scale 1:100.000, while our tiles go
> to 1:1000 and below (z19), this is the problem why the CLC is not suitable
> for OSM, but it is perfectly OK to be viewn in say zoom8 or 9, maybe 10, but
> these are not the zoomlevels in which we usually look at our data.

CLC could have been used in a much more sensible way if the import
wouldn't have been just copy-pasting it fully to osm. CLC is somewhat 
reasonable only for a very large, continuous body of something (although 
the classification would still be occasionally wrong but that's manually
fixable afterwards). Trying to get any small feature out of CLC is/was
definately an error and effectively just importing noise.

A more clever approach would have filtered out all small features and 
shrinked boundaries by some amount to avoid boundary overshoot errors. 
That would have allowed big areas of homogeneous landuse to be imported 
with a rough outline that is not very likely to collide with the nearby 
landuse features. This would have been kind of low hanging fruit only 
approach. Obviously it would have also left some gaps to annoy mappers
to fix the outlines ;-).

-- 
 i.