[OSM-talk-be] Hi all

[Kenny Knecht]

2010/1/6 Lennard <ldp at xs4all.nl>

> > but I think the NGI won't even consider doing that. In the import
>
> It's certainly worth it to ask and explore this, before we turn all our
> attention to CLC2000.
>
>
I have contacted them: let's see what comes of it...


> > section they describe how to do it.
> > Do you all think it is worth it?
>
> I have the entire CLC2000 dataset on my server, and had a quick look. The
> answer is, obviously: "yes and no".
>
> Con: The data is old, some things will simply be outdated.
>
> Pro: Some objects, like residential landuse areas, look pretty good. All
> the typically belgian 'linten' are very recognisable.
>
> Con: Other objects, like agricultural areas, are pretty large, and would
> certainly not form a single continuous area IRL. This would need extra
> checking after the import.
>
> Pro: landuse mapping in Belgium, outside where there is yahoo hires
> coverage and people have spent time tracing, is pretty lacking. Taking the
> specific things from CLC2000 that would work nicely would be a good step
> forward.
>
> Pro: the way the French have setup the Corine import is that you can
> exclude areas which are overlapping with existing OSM areas. You can then,
> later on, import those areas missed due to overlap, by hand, from a
> website. Ben said we could even delete existing OSM areas before the
> import starts, if we know the existing mapping is inaccurate.
>
> If you guys want, I could set up a map overlay with the raw CLC2000 data,
> so you can check it out?
>

The map overlay would be handy.
Appartenly the corine dataset consists only of chuncks that are at least
25ha in size: that means that you cannot expect too much detail out of it,
which in turn favours the fact that we should retain the manual data above
Corine data, but on the other hand makes the dataset robust towards changes,
especially in Belgium, so I guess that the changes between 2000 and 2006
will not be dramatic.



>
> > I have got some more plans, but I promise I want do any bulk uploads
> > without consulting you ;-)
>
> Do you have any other potential datasets in mind?
>

Another example is the clean-up of the place-key. Right now this is a mess:
some have postal_code attached some have not, some have an is_in tag some
not.
Right now I have downloaded all the place-key in belgium (which involved
downloading all the boundaries of belgium and testing whether the places I
got were in that disconnected polygon :-o )
On
http://www.post.be/site/nl/residential/customerservice/search/postal_codes.htmlyou
can download postal code of municipalities and sub-municipalities per
province,  which gives you a nice hierarchical overview of the Belgian state
structure. This will at least allow me to
- create a list of the missing municipalities and sub-municipalities (from
now on called muns)
- create a list of the unknown or misspelled muns in OSM
- create consistent information for all of the muns which are already
present in OSM (eg population for muns Flanders can be found on
inwoners%20gemeenten.xls<http://www.binnenland.vlaanderen.be/statistieken/excel/inwoners%20gemeenten.xls>)

The question is what do we do with muns not present in OSM?
There are a number of sources we could use to generate them from:
- The geonames from the US military databases which have no copyright
attached *www.nga.mil*. This has been used previously in OSM.
- another one is this link
http://www.rdlt.com/postcodes-van-alle-gemeentes-van-belgie-met-gps-coordinaten
But where does the data come from? Can we use it?
- On the website http://www.hisgis.be/ you can find detailed shp files,
which are copy-righted but I think we can use them to curate the data in one
or both of the previous datasets. Since we only use the data but not publish
it, I think there is no legal problem.

We must also think about the nomenclature
If you look at http://tagwatch.stoecker.eu/Belgium/En/keystats_place.htmlyou
see that there are 321 towns and 1642 villages. There are even 14
cities, which means according to
http://wiki.openstreetmap.org/wiki/Key:place  +100000 inhabitants!
Personally I would ignore that guideline completely because it is solely
based on # of inhabitants. As is clearly explained in
http://en.wikipedia.org/wiki/List_of_municipalities_of_Belgium,
municipalities and sub-municipalities reflect the Belgian state structure
more closely. We could of course map city to muns > 100000, towns to other
muns and villages to sub-muns, but that will not be true always (i.e.
Herstappe is a mun with 84 inhabitants!)
But that is my opinion: this issue is certainly open to discussion.

That way I hope to get the place-key curated soon. If you have any remarks
about methodology or sources, please feel free to email. Is there a server I
can put files, programs and such on? That way discussion might be more
concrete...

So long,

Kenny aka Pietje Leugenaar
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-be/attachments/20100108/fd114337/attachment.htm>