[OSM-talk] new mailing list request - OSM outdoor/natural phenomena mapping

Richard Z. ricoz.osm at gmail.com
Wed Mar 5 19:41:52 UTC 2014


On Wed, Mar 05, 2014 at 04:04:10PM +0100, Martin Koppenhoefer wrote:
> 2014-03-05 14:42 GMT+01:00 Richard Z. <ricoz.osm at gmail.com>:
> 
> > As an example, we are having repeated discussions how to tag forrest but
> > did not even start thinking about a generic concept how to map vegetation
> > such as:
> >
> > climatic zones
> > vegetation zones
> > soil biology
> > vegetation layers
> >
> 
> 
> 
> most of these do not seem remotely suitable for the osm data model and the
> way we collect and store data. Climatic zones and vegetation zones are very
> big areas with fuzzy boundaries. We do not even manage to map huge areas
> with clear boundaries (e.g. look for the atlantic ocean in osm), how could
> we start to map those with fuzzy boundaries?

the zones may not be as huge as you would imagine. In some cases the zones are
just square miles or much smaller with very sharp boundaries as seen on Mauna Kea,
Haleakala, or some places on Kauai.

http://upload.wikimedia.org/wikipedia/commons/b/bb/Koppen_World_Map_%28retouched_version%29.png

And if we have problems with the Atlantic Ocean we should fix them:)

> These also might be subject to
> generalization and interpretation to a degree that 2 scientists in the same
> field would draw the borders differently.

this could happen, the bigger problem I see is that in every country they would
apply the same rules slightly differently. 
But we have similar problems in most other domains and still try to map them. 
Just recall the discussion of highway=track classification.

> IMHO these could maybe produced
> out of osm data (together with other data like precipitation, temperature
> etc.) 

interesting idea. You would also need a very good elevation model and prevailing 
winds, soil properties and maybe some other details of course.
However I believe that a climatic zone is an empirical data set and trying to 
derive it from other information is like trying a 4 weeks weather forecast. It may
work in some areas but the result will be worse than using observed data.

Another point is, the climatic zones may be useful to predict vegetation
characterstics for large areas of the world where detailed vegetation
mapping is not available yet.

> if you'd analyzed the occurence of certain species (the tags for
> mapping single plants are there, you only have to use them, currently these
> are the numbers:
> 319 553
> *species* <http://taginfo.openstreetmap.org/keys/species>
> 125 867
> *species*:de <http://taginfo.openstreetmap.org/keys/species%3Ade>
> 81 881
> *species*:it <http://taginfo.openstreetmap.org/keys/species%3Ait>
> 
> 131 887* taxon* <http://taginfo.openstreetmap.org/keys/taxon>

I think that really underscores the need for a mailing list for such subjects
because I had never the idea that this tags exist.
 
> Tags regarding soil biology are also very hard to verify on the ground by
> the general mapper. They require specialized knowledge and maybe also a
> laboratory to do analysis and similar. 

frequently people searching mushrooms will have very good knowledge 
of this subject. I don't hope to get a perfect map of soil biology
anytime soon but having a framework for it prepared should not hurt.

> Even if collecting the data wasn't
> an issue (say it would be possibile to import "perfect" data), still it
> won't integrate or fit well with the datamodel (this is more statistical
> data than actual hard facts, and drawing the border is almost impossible).

I find it always hard to decide where to draw borders of forrests because they 
are fuzzy and unfortunately many lakes and rivers have shores which are changing 
very quickly over time which is an even bigger problem.

> Your last point, vegetation layers, might be a little bit different. IMHO
> this could be done, and partly it already is (see the landcover-key)

yes, landcover is a step in the right direction but IMHO nowhere close
to a satisfactory mapping of vegetation and ground properties.
Mapping for example the Sonoran desert would require yet another approach,
it is essential to specify the properties of the partially exposed ground 
as well as the vegetation.

Richard



More information about the talk mailing list