[Imports] Hungarian CLC import

[Martin Koppenhoefer]

2012/10/29 Balázs Szalkai <bszalkai0 at gmail.com>:
> More than about one third of Hungary has already been imported from the CLC
> dataset by others. The following tag is used for attribution:
> source=© EEA, Koppenhága (2009); Készítette a FÖMI a KvVM megbízásából
> (2009)
> I'm not a lawyer but I think this tagging is correct. see
> http://lists.openstreetmap.org/pipermail/talk/2011-April/057501.html


I think the question about tagging was refering to how you translate
the Corine-tags to OSM-tags, and which tags from Corine you translate
and which you drop. I am also interested in the answers to this.



> What I do about existing data is: I check all the existing landuse polygons
> by hand before import. I remove most of the landuse=residential polygons, as
> they are usually also present in the CLC dataset and the CLC versions are
> more accurate (checked with Bing).


this is surprising news, really, CLC is more accurate in your area
than what is in OSM? In the past I checked some places in France and
Italy and CLC was never almost as accurate than OSM, in fact in Italy
the quality was that bad that we decided not to import it at all
(given that CLC is a medium resolution dataset, at least in Italy, it
was not surprising that it wouldn't fit well with OSM, which is high
resolution). These are all observations from different countries and I
am not saying that in Hungary the situation might not be different.


> There are also some existing sporadic
> forests and meadows, but usually they are also represented in CLC so then I
> delete them and copy the extra information (e.g. name of forest) into the
> newly imported forest polygon.


You shouldn't really do this, as it is mostly considered bad practise
or even vandalism, as you are deleting the history of these objects.
Usually what is already there should have precedence over an import
(and should be carefully considered also when working manually).


> There were 2-3 cases when CLC and existing
> data disagreed about a forest, CLC marked it as farm. Then I left both the
> existing and new data in place and tagged the forests with "fixme=please
> check this, i don't see any forest here on Bing and CLC marks this as farm,
> too" or something.


If there is no forest, you don't have to keep it, but to be quite sure
(without knowing the place) of course depends on the age and
resolution of your aerial imagery (e.g. around here in the countryside
there are Bing aerials from last year in quite decent resolution, and
if there is farmland on these, there likely won't be a forest in
reality), so when in doubt I agree it is better to keep what is
already there in OSM.


> I don't think CLC is low resolution. Vineyards, meadows, forests, etc. are
> usually accurate. See for example
> http://www.openstreetmap.org/?lat=46.3314&lon=19.0919&zoom=13&layers=M You
> can see here that CLC is well-detailed.


Yes, it is detailed, but there is not a single edge coinciding with
the aerial imagery. If you traced these features by hand from aerial
imagery, would you do it in the way CLC did? Despite the resolution I
still noticed quite a lot of very close nodes and strange corners
(IMHO data errors) and some middle nodes in straight segments. It
further seems all polygons are overlapping and there are no
multipolygon-relations at all? This makes further refinement quite
hard and is also a lot of redundacy which could be avoided with
relations.


> Nevertheless, there are some voids
> that need to be filled manually in the future, as you can also see on the
> linked map fragment.


IMHO what you can see in this area needs heavy work in the future, not
just the gaps.

cheers,
Martin