I don’t understand “less than 1:100,000”. Resolutions will be in metres, not pure numbers, and will depend on the source of the data, , and if the data was only available in bit map format, the accuracy of the tracing and the degree of approximation used to keep the number of line segments manageable. For US states, they are likely to be very accurate, because they boundaries are mainly straight lines.
Apologies, i am using the word ‘resolution’ when I should probably be using ‘scale’.
These are admin boundaries I am referring to so I doubt they are from bitmap sources and/or traced. I would assume they are from various national government sources.
So the original data would, I believe, be vector data and would have been supplied at a certain scale i.e. accuracy.
If I overlay some German boundaries from an online source which are at 1:100,000 scale over an OSM tile layer they do not match the boundaries shown on the OSM tiles - OSM is much more accurate/detailed. So I am assuming the OSM source was from a larger scale e.g. 1:10,000 perhaps.
But what I would like to know is what that original scale was before being rendered into OSM tiles?
Asking for scale or resolution of OSM data is not usually very useful. This is because the data is created by tens of thousands of people using a vast multitude of data sources.
In heavily mapped areas OSM data is typically 80% within 5 m of data from ‘authoritative’ sources (e.g., national mapping agencues), and about 95% within 10 m. This is true, for instance, of the road network for most of Europe. That being said many waterways in the UK are as much as 200 m out of position (largely because they were mapped from earlier lower resolution sources). Ultimately I would hope we get most OSM data to this level of accuracy: roughly corresponds to 1:10k or 1:25k scale mapping. Further positional accuracy is beyond our current techniques, although topological accuracy should be OK.
Exactly the same applies to boundary data, sometimes it has been imported from another source, such as a national mapping agency, but other times it is built progressively by iterative refinement. Thus a single country border may have elements which are accurate to within a metre compared with government sources and others which are 10s, 100s or 1000s of metres out.
To give a practical example from Ireland. Until last year some county boundaries were extremely crude (notably in Munster) with the mapped ways being several kilometres distant from the their correct locations. On the other hand parts of Northern Ireland had a much higher degree of accuracy at the same time. See my blog post for an analysis of accuracy in Northern Ireland: http://sk53-osm.blogspot.co.uk/2015/12/how-accurately-have-townlands-in.html.
Internal boundaries in Ireland are now largely accurate, the remaining significant source of inaccuracy probably lies in the coastline. This will be true for many countries: particularly for parts of the coastline in rural or uninhabited areas.
All of this being said, it is my own experience that OSM boundaries are more useful for anything requiring greater precision than the data available from Natural Earth (which at best is generalised for 1:10M scale). I would have high confidence for data for most European countries, slightly less for US/Canada (because there were some projection conversion issues in original data). For the rest of the world I dont have the same knowledge, but I would expect borders to be accurate in populated areas and less so elsewhere (e.g., in mountain ranges, such as Nepal/China).
Nowbody knows that. It depends on the origin data, the mapper(s) did use. And how much he/they generalized during inputing those ways. This may be different for every boundary way - even in the same country. No mapper mapped the complete boundary, it’s team work.
And if a mapper is e.g. located in Belgium and he has very good legal data from his goverment, this area may be better then other parts of the germany boundary…