search

AndGem / osm_extract_polygon

Extract Osmosis compatible and GeoJson polygon files from OSM PBF files
MIT License
17 stars 3 forks source link

Partial polygons #51

Open Gared opened 1 month ago

Gared commented 1 month ago

I don't know if this an issue with this library or osm data in general: For example if I try to extract germany-latest.osm.pbf it extracts also borders for austria, switzerland, etc. with admin_level=2, but the polygon in it is only the border between germany and austria and not the whole border of austria. Is there a way to extract only "whole polygons"?

AndGem commented 1 month ago

Hey @Gared ,

I looked a bit around, but it looks really like a data problem. I found for example this relation: https://ra.osmsurround.org/analyzeMap?relationId=51239

It is indeed incomplete.

However, in theory there might be a way to actutally still do it by merging multiple relations at the right places (and potentially interpolate a bit. This will be a bit more tricky. I'll play around with this in the next days to see if I can manage something (because my original intention was also to have this option.. I never tried it on this admin level and did not notice these kinds of problem, so thanks for reporting it).

AndGem commented 1 month ago

As I feared, the data makes this really difficult. I had hoped that I can find overlapping nodes that I can use to reconstruct the whole border by stitching individual OSM relations together. However, already this doesn't work for germany-latest that I downloaded a few days ago.

I don't have any good idea how to do this in a reliable way. If you have any idea, I'm happy to try it out though.

morandd commented 1 month ago

this is pervasive.... the GeoFabrik "germany-latest.osm.pbf" etc files are cut harshly and include slivers of neighbouring countries. Removing these slivers is tedious. And this cannot be automated since how can you differentiate a sliver from a small, valid region?

What someone (who? a nice volunteer?) needs to do is add the country name or parent relation ID to every single admin_level relation in OSM. Or produce a blacklist of these slivers which should be removed from each GeoFabrik country. Or, GeoFabrik should use a gentler cutting technique when building the "[country]-latest.osm.pbf" files that doesn't introduce the slivers.

Clean boundaries are available: Google Maps, https://osm-boundaries.com/, maybe Overture Foundation, GDAM (but no commercial use), and some commercial providers like ESRI, MapBox, and MB International.

This issue is acute in Europe due to many countries with complex borders. But it occurs globally. If you are focused on Europe a solution is to use NUTS/LAU instead.