Re: [OSM-fi] Fixing broken and old-style multipolygons
I'm not sure but I think we have the whole country laser scanned and the huge data available from MML. Maybe someone skilled enough could write a forest detector and import the forest data from there? Just my 2 cents. 3.4.2017 3.21 ip. "Jochen Topf" kirjoitti: > On Mon, Apr 03, 2017 at 02:13:23PM +0300, Jukka Rahkonen wrote: > > Right, those seem to come from the notorious Corine Landcover import that > > was never completed nor reversed even it does have loads of topological > > errors. It does bring some colour to the map in about 25% of the Finnish > > territory, though. You may know that we do not have very good alternative > > sources for land cover imports because 71,6% of the area of Finland is > > covered by forests. Therefore our topographic map database does not even > > have a class for forests - they are background, not interesting at all. > > > > I had a look at some polygons and I do not know exactly how they should > be > > corrected. For example relation > > https://www.openstreetmap.org/relation/1481481#map=8/62.740/25.565 > > > > It has an inner member https://www.openstreetmap.org/way/104644001 which > > somehow looks the same as > > https://www.openstreetmap.org/way/104646035 > > > > The inner polygon, that is coniferous wood, is a simple polygon with only > > outer ring. So as a stand-alone it could carry its own tags. But because > the > > ring is re-used as a hole of the bigger mixed wood polygon it should not > > have tags. Instead the single ring coniferous area should be changed > into a > > multipolygon relation that has only one member, and that new relation > would > > get the coniferous tag. Is this correct? > > No, simple closed ways as polygons are okay. The problem in this case is > that the relation has one outer ring and three inner rings, all of which > have the same tags (natural=wood, leaf_type=mixed). This is a not > uncommon sub-problem of old-style tagging and is not allowed any more by > new-style tagging. We would interpret this now as additional polygons > filling > the holes with exactly the same type of landcover the outer polygon has, > which, of course, doesn't make much sense. To fix this the tags from the > inner and outer ways have to be removed and put on the relation. There > might or might not be a different closed way or multipolygon relation > covering the whole to get some different type of landcover. If there is > a different way that covers the inner, it could be removed and the inner > tags can be put on the way that is the inner ring. > > > I really much doubt that anybody is interested in correcting the Corine > > polygons but people has not been too eager to revert the import either. > > Because of the colour, you know. But let's see how other members of the > > Finnish community react. I fear also that OSM will never get rid of the > > hand-written multipolygon relation system and because the reason for the > > trouble will not be fixed you have been forced to start your huge effort > > with fixing the broken data > > https://wiki.openstreetmap.org/wiki/Area/The_Future_of_Areas. Good luck > for > > you. > > We are currently not looking at changing anything in the basic OSM data > model, but osm2pgsql software *will* change in the not too far future and > this *will* have a visible impact on the map in Finland if nothing is > done. You can see what these changes will bring by looking at this map: > > http://area.jochentopf.com/map/index.html#8/61.967/27.554 > > Switch off the overlay and move the slider around and you'll see that a > lot of the green will be gone. There will also be areas that don't show > up as forest now, but will in the future. So if nobody cares about this > issue, the map will look even more wrong in the future than it is now. > > I hope the Finish community will decide on some plan for action here. I > am sure the global OSM community will help to the best of their ability, > but doubt many people want to spend their time fixing up imported data > that isn't that high quality anyway. > > Jochen > -- > Jochen Topf joc...@remote.org https://www.jochentopf.com/ > +49-351-31778688 > > ___ > talk-fi mailing list > talk-fi@openstreetmap.org > https://lists.openstreetmap.org/listinfo/talk-fi > ___ talk-fi mailing list talk-fi@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-fi
Re: [OSM-fi] Fixing broken and old-style multipolygons
On Mon, Apr 03, 2017 at 02:13:23PM +0300, Jukka Rahkonen wrote: > Right, those seem to come from the notorious Corine Landcover import that > was never completed nor reversed even it does have loads of topological > errors. It does bring some colour to the map in about 25% of the Finnish > territory, though. You may know that we do not have very good alternative > sources for land cover imports because 71,6% of the area of Finland is > covered by forests. Therefore our topographic map database does not even > have a class for forests - they are background, not interesting at all. > > I had a look at some polygons and I do not know exactly how they should be > corrected. For example relation > https://www.openstreetmap.org/relation/1481481#map=8/62.740/25.565 > > It has an inner member https://www.openstreetmap.org/way/104644001 which > somehow looks the same as > https://www.openstreetmap.org/way/104646035 > > The inner polygon, that is coniferous wood, is a simple polygon with only > outer ring. So as a stand-alone it could carry its own tags. But because the > ring is re-used as a hole of the bigger mixed wood polygon it should not > have tags. Instead the single ring coniferous area should be changed into a > multipolygon relation that has only one member, and that new relation would > get the coniferous tag. Is this correct? No, simple closed ways as polygons are okay. The problem in this case is that the relation has one outer ring and three inner rings, all of which have the same tags (natural=wood, leaf_type=mixed). This is a not uncommon sub-problem of old-style tagging and is not allowed any more by new-style tagging. We would interpret this now as additional polygons filling the holes with exactly the same type of landcover the outer polygon has, which, of course, doesn't make much sense. To fix this the tags from the inner and outer ways have to be removed and put on the relation. There might or might not be a different closed way or multipolygon relation covering the whole to get some different type of landcover. If there is a different way that covers the inner, it could be removed and the inner tags can be put on the way that is the inner ring. > I really much doubt that anybody is interested in correcting the Corine > polygons but people has not been too eager to revert the import either. > Because of the colour, you know. But let's see how other members of the > Finnish community react. I fear also that OSM will never get rid of the > hand-written multipolygon relation system and because the reason for the > trouble will not be fixed you have been forced to start your huge effort > with fixing the broken data > https://wiki.openstreetmap.org/wiki/Area/The_Future_of_Areas. Good luck for > you. We are currently not looking at changing anything in the basic OSM data model, but osm2pgsql software *will* change in the not too far future and this *will* have a visible impact on the map in Finland if nothing is done. You can see what these changes will bring by looking at this map: http://area.jochentopf.com/map/index.html#8/61.967/27.554 Switch off the overlay and move the slider around and you'll see that a lot of the green will be gone. There will also be areas that don't show up as forest now, but will in the future. So if nobody cares about this issue, the map will look even more wrong in the future than it is now. I hope the Finish community will decide on some plan for action here. I am sure the global OSM community will help to the best of their ability, but doubt many people want to spend their time fixing up imported data that isn't that high quality anyway. Jochen -- Jochen Topf joc...@remote.org https://www.jochentopf.com/ +49-351-31778688 ___ talk-fi mailing list talk-fi@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-fi
Re: [OSM-fi] Fixing broken and old-style multipolygons
Hi, Right, those seem to come from the notorious Corine Landcover import that was never completed nor reversed even it does have loads of topological errors. It does bring some colour to the map in about 25% of the Finnish territory, though. You may know that we do not have very good alternative sources for land cover imports because 71,6% of the area of Finland is covered by forests. Therefore our topographic map database does not even have a class for forests - they are background, not interesting at all. I had a look at some polygons and I do not know exactly how they should be corrected. For example relation https://www.openstreetmap.org/relation/1481481#map=8/62.740/25.565 It has an inner member https://www.openstreetmap.org/way/104644001 which somehow looks the same as https://www.openstreetmap.org/way/104646035 The inner polygon, that is coniferous wood, is a simple polygon with only outer ring. So as a stand-alone it could carry its own tags. But because the ring is re-used as a hole of the bigger mixed wood polygon it should not have tags. Instead the single ring coniferous area should be changed into a multipolygon relation that has only one member, and that new relation would get the coniferous tag. Is this correct? I really much doubt that anybody is interested in correcting the Corine polygons but people has not been too eager to revert the import either. Because of the colour, you know. But let's see how other members of the Finnish community react. I fear also that OSM will never get rid of the hand-written multipolygon relation system and because the reason for the trouble will not be fixed you have been forced to start your huge effort with fixing the broken data https://wiki.openstreetmap.org/wiki/Area/The_Future_of_Areas. Good luck for you. Regards, -Jukka Rahkonen- Jochen Topf kirjoitti 2017-04-03 13:25: Hi! You might have seen that a few weeks ago I started a huge effort to clean up broken (multi)polgyons and old-style multipolygon relations (with tags on ways instead of on the relation) in OSM. You can find out more about this project on http://area.jochentopf.com/ . I am writing on this list here, because Finland is a major hotspot for those multipolygon problems. See these maps: http://area.jochentopf.com/map/index.html#7/61.923/25.378 http://tools.geofabrik.de/osmi/?view=areas&lon=28.31590&lat=62.06443&zoom=8 This is probably due to some imported data. Is anybody here interested in this topic? How can we move this forward? Jochen ___ talk-fi mailing list talk-fi@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-fi
[OSM-fi] Fixing broken and old-style multipolygons
Hi! You might have seen that a few weeks ago I started a huge effort to clean up broken (multi)polgyons and old-style multipolygon relations (with tags on ways instead of on the relation) in OSM. You can find out more about this project on http://area.jochentopf.com/ . I am writing on this list here, because Finland is a major hotspot for those multipolygon problems. See these maps: http://area.jochentopf.com/map/index.html#7/61.923/25.378 http://tools.geofabrik.de/osmi/?view=areas&lon=28.31590&lat=62.06443&zoom=8 This is probably due to some imported data. Is anybody here interested in this topic? How can we move this forward? Jochen -- Jochen Topf joc...@remote.org https://www.jochentopf.com/ +49-351-31778688 ___ talk-fi mailing list talk-fi@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-fi
