[Tagging] Using multipolygons to map bays in Alaska
kevin.b.kenny at gmail.com
Thu Nov 15 18:46:46 UTC 2018
On Thu, Nov 15, 2018 at 12:16 PM Christoph Hormann <osm at imagico.de> wrote:
> > I'm afraid that I'm not following this argument very well. What about
> > a bay is 'completely non-verifiable?'
> The geometry.
> These geometries:
> are completely non-verifiable. They add data but they does not add any
> substantial information about the verifiable geographic reality to the
> database that could not be represented with a single node.
We agree that those are mismapped examples. If you're going to map a bay
as an area, extend its boundaries to the shore!
With these geometries:
> a large portion of the geometry and as a result the derived way_area are
> completely non-verifiable. Also here a properly placed node would
> together with the coastline transport all the verifiable information
> about the geographic reality there is.
Even in that extreme example, having the spatial extent adds value.
Consider a large-scale map of the villages of Audierne
https://www.openstreetmap.org/node/26698453 and Poulgoazec, rendered on a
sheet of paper. How will the renderer of such a paper map determine that
the body of water on the southern edge of the sheet is to acquire the name,
'Baie D'Audierne', without information about the spatial extent of the bay?
(I do not ask this question in terms of any particular piece of rendering
software - and in particular not Mapnik-OSM Carto. The question is how,
even in theory, any conceivable renderer can be expected to make that
Does the same argument apply for an indentation in the coastline that's
more nearly totally enclosed? Is the Sea of Azov
https://www.openstreetmap.org/relation/5486417 mapped correctly? The Black
Sea? https://www.openstreetmap.org/relation/7160849 Would the Adriatic Sea
deserve such treatment (right now it's just 'coastline' and maybe a node
that I can't find). The Ægean Sea? The Tyrrhenian Sea? The Ionian Sea? The
Gulf of Taranto? (Moving out to lesser and lesser amounts of enclosure.)
Closer to home for me, would Jamaica Bay (the area seen in
https://www.openstreetmap.org/#map=12/40.6007/-73.8731) rate a
multipolygon? RIght now it's just coastline, which leads it to be accounted
as 'Atlantic Ocean'. That would puzzle the locals, who often refer to 'the
ocean side' and 'the bay side' of the Rockaway Peninsula. Here, too, the
extent of the bay is pretty well determined except for the precise location
of its mouth. (That one might most 'objectively' be placed at the narrowest
part of Rockaway Inlet, under the Marine Parkway Bridge, but local mariners
instead say that one enters Jamaica Bay when one crosses the imaginary line
between the lighthouse on the Rockaway Point jetty and the light on
Steeplechase Pier at Coney Island, and I generally defer to the locals when
it comes to knowing the extent of a place.) If I were to produce a
large-scale paper map of my former home town of Inwood
https://www.openstreetmap.org/relation/174930, without spatial information,
how would my renderer know that the large water body at the left edge of
the sheet should be labeled, 'Jamaica Bay' or that the small one that would
likely appear on the sheet's southwest corner is Norton Basin? With partial
areas that are as represented as [multi]polygons, the rendering toolchain
that I use can handle the labels quite nicely today. I have absolutely no
idea how I'd make it go searching for a nearby node, or what to do when if
Is the answer that, "if you want to produce such a map, OSM is the wrong
tool for the job'? Or, 'well, given just the node, eventually a
sufficiently sophisticated AI will be able to reconstruct the area?'
Neither answer helps very much with the practical problem, although they
would at least inform me that I need to maintain hydrographic data separate
from OSM at least until such time as the miraculous AI arises.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Tagging