|
Hi folks! Is there a way to get all coastlines (consistent (!) polygons that define landmasses) out of OSM data? Like GSHHS, but the detailed data that OSM offers. Do you know of a data provider that offers these? I checked a few shapefile exports available on the net, but they seem to be polylines that sometimes aren't closed. Thanks in advance, Bernd |
|
First of all you need to understand that OSM is a volunteer operated project and that at any given time there are very likely to be errors - e.g. holes in the coastline, missing islands, even "flooded" parts of continents - in our dataset. This is something you have to cope with when using OSM data. The openstreetmapdata.com web site is probably the nearest you will get to a bug-free landmass polygon dataset. You have a choice of large landmass polygons, or smaller overlapping ones. The software used to produce these data sets tries to fix smaller problems (small gaps etc.) automatically. |
|
Of course, most of the OSM mapping users, pretending on a vector (transmission) based rendering are doing this. The land/sea layer is, as a rule, the basic (or the first) layer to render. But there are various models to convert the source coastline poly-lines to a land/see area (call them multi-polygons if you like) presentation/coverage (or to a defragmented “landmasses” as you call them). We/I perform this conversion frequently on the Wednesday/Thursday OSM dump including procedures like: LAtLon to Mercator; eliminate replicated poly-lines, vectors, nodes; poly-lines to polygons; error detection and reparation; polygons to (simple) areas and the like. An example for Europe (extended, because of the Spanish islands) “landmasses” from some weeks ago you may download (and examine) from here: https://docs.google.com/folder/d/0B6qGm3k2qWHqZ2N3RkdUbnQwem8/edit This land geometry is the input to a scale/zoom levels’ generation to provide data reduction for a continuous scaling. An example of these is also available as land_Lev7 from the same place. This level become active from around 1:50 000 000 scale and it contains 1 464 nodes compared to 6 179 103 nodes in the source. If you overlap the two geometries in the mentioned scale (eventually smaller) you will not see any differences except the better contrast and readability for the pre-scaled version. Finally, if you zoom in to around 1:10 000 on the land geometry around (137 862;6 782 060) position you will see a typical logical error (compared to formal errors). There are many tiny channels and river/fiord systems tagged as coastline. These should logically be in other area object classes. |
