[NUUG kart] [Imports] kartverket imports to OpenStreetMap

[Christoph Hormann]

On Wednesday 16 October 2013, Tyrfing OSM wrote:
> > [...]
> >
> > But it
> > is unavoidable due to the sheer amount of data you add that there
> > are mechanical decisions made in the process that would deserve
> > being discussed.  This is what makes your edits different from use
> > of Bing imagery as source.
>
> I always appreciate comments on my use of tags, that's the best way
> to improve. But I would tag the same way if I used a different source
> or if the changesets were smaller.

I find this unlikely - take the width tag for the streams for example - 
of the 952434 waterways in the kartverket data set 841610 have 
a 'vannbr' or 2, the rest 3, 2 is used for waterways hardly visible in 
Bing images (i.e. < 1m) up to at least about 5m width.  Quite obviously 
this attribute does not (at least not properly) indicate the width in 
meters, especially since it would not make any sense not to have any 
waterways less than 2m wide in a 50k topographic map.  You would not 
tag these values if you were mapping based on Bing images or local 
knowledge.

The production of the kartverket data involved classifying features into 
a fixed set of categories and you make use of these classifications in 
your tagging.  This is one of the things that need discussing with 
imports.  As said the usual first step would be to document the 
attributes in the source data and their meanings on the wiki and your 
plans how to translate them into tags.  This is currently incomplete.  
In case of the waterways for example there are two more attributes not 
currently documented in the wiki where the meaning eludes me:

malemetode [various values between 18 and 92]
noyaktighet [100..5000 in certain intervals]

Those might be helful to select proper tags for the elements.

You can also think of it this way: Aerial images and in-the-field 
surveys are primary data sources, there are very few subjective 
decisions already predetermined in that data.  This is different for 
maps created outside of OSM where a lot of decisions how to interpret 
the primary data have already been made outside the influence of the 
mapper using the data.  The mechanical edits policy applies to edits 
were the mapper does not have full control over all important aspects 
of the process from primary source to the upload.

> I agree that a river is a large stream. To classify a stream as a
> river one might consider other aspects than merely the width. E.g:
> Length, average discharge (volume rate of water flow), size of the
> drainage bassin, how easy it is to cross (which would depend on
> discharge and width), local classification/name or relative size
> compared to other streams in an area (which then of course might
> depend on any of the other aspects). Also the width (and depth, it
> that was to be considered) tend to vary along the way.

Yes, the current stream/river distinction is not very useful from a 
cartographic viewpoint - i have been saying this for quite some time.  
It is however difficult to think of a practically verifiable criterion 
other than width for making this distinction (The wiki article on 
verifiability http://wiki.openstreetmap.org/wiki/Verifiability 
specifically uses the river/stream distinction as an example).

It is best not to think of stream/river as a distinction to be used for 
showing/hiding waterways in map rendering (although it is widely used 
that way of course).

> I've noticed that somebody recently added a rule to distinguish
> waterway=river from waterway=stream on the wiki page for stream, but
> I'm not sure how/where that rule (which is related to how far an able
> person can jump) became commonly accepted?

I added this to that wiki page but it has been around for a long time 
(on http://wiki.openstreetmap.org/wiki/Key:waterway since 2007).

> > You also seem to tag all wetlands with wetland=bog but i am
> > not sure if this is correct in all cases, especially with wetlands
> > around a river.  But i lack the local knowledge to tell if this is
> > truely wrong in the individual case.  Again something that would
> > need to be discussed if such tag is applied generally to all
> > wetlands and is not decided individually.
>
> The wetlands I've edited are mire (No: "myr"). These are either bog
> or fen, and the OSM tagging scheme seem to use bog for both. So I
> still think those are correct.

Bog is usually only used for wetlands producing peat.  If the 
classification of 'myr' in the kartverket data refers to the same would 
need to be determined.  Unless there are other wetlands in the data set 
which are not 'myr' this is doubtful though.  Bogs are widespread in 
Norway but certainly not the only kind of wetland that exists there.

> > Note a thorough import discussion could also help with your
> > workflow - the lack of orientation in the waterways you fix by hand
> > for example can probably mostly be corrected automatically for the
> > whole data set based on the elevation data sets kartverket also
> > made available. [...]
>
> Yes, I assume an import also could create multipolygons, remove
> double nodes etc. I'm sceptical to imports because I struggle with
> the results of them quite often (like the Corine disaster). Currently
> I'm happy doing the edits manually, thus improving the map slowly
> (compared to an import).

In commonly used terms the distinction you make here is between a well 
made import and a bad import.  Good imports in areas with significant 
data already there are slow - no matter if you automatically create 
multipolygons etc. or not.

Greetings,

-- 
Christoph Hormann
http://www.imagico.de/