[Imports] Serbia boundaries

Branko Kokanovic branko at kokanovic.org
Sat Jan 18 22:17:55 UTC 2020

Announcing here that we plan to add some tags to Serbian boundaries on admin_level=9[1] as Serbian community received open data set of boundaries from government[2] (all admin levels).

We did some preliminary analysis and this government data seems really good. It also turns out we mostly have all boundaries with good enough accuracy[3]:) (average of 95% of common boundary area between government and OSM data) Because of this, at this moment we do not plan to improve existing geometries more (but see "open question" at the end of mail), and we decided to preserve existing history, so we also do not want to delete existing ways and import it from scratch. However, we did wrote procedure for manually fixing existing ways[4] to match government boundaries if we want it to be 100% correct.

So far, plan is only to add new tag `ref:sr:maticni_broj` (similar to de:regionalschluessel and de:amtlicher_gemeindeschluessel) which is official reference in cadaster of districts/municipalities/cities/settlements and to improve subarea role (add where missing). We will ensure both of these are in place by leveraging existing Serbian lint tool[5] (by emitting warning if boundary don't have `ref:sr:maticni_broj` or subarea).

OSM file with all admin_level=9 boundaries prepared from government data is here [6]. We are actively working in open, in this repo[7] for all current and future analysis about this data and potential automated conflations. All of our decisions are discussed in forum here[8].

Licence is fully open for "all purposes" and opening data is ratified in parliament in December, and present in official journal (Сл. гласник РС 86/2019). It can be found (for this and other Serbian government data) at [9].

We *do not* plan to touch any country boundaries, just internal ones (although, plan is to contact bordering countries' forums, just to see how much we do differ based on their open data, if any).

**Open question**: if anyone can work with us and help how to conflate existing OSM ways shared between admin_level=9 boundaries, so they are aligned better with produced .osm file[4] (government data), any tips, resources, what to watch out... please shout out! Current 95% accuracy that we have is good, and writing tool to "move"/"add" nodes of boundary ways while watching out for various edge cases is possible, but not really justifiable effort. If there is existing solution/half-solution in the wild that we can take and adopt, it might be a better way forward!

Thanks, Branko

[1] https://wiki.openstreetmap.org/wiki/Serbia/RGZ_Import
[2] https://opendata.geosrbija.rs/
[3] https://kokanovic.org/rpj-analysis.csv
[4] https://wiki.openstreetmap.org/wiki/Serbia/RGZ_Import/Uputstvo (in Serbian)
[5] https://gitlab.com/stalker314314/serbian-osm-lint
[6] https://kokanovic.org/rpj-v1.7z
[7] https://gitlab.com/stalker314314/prostorne-jedinice-import
[8] https://forum.openstreetmap.org/viewtopic.php?id=68288
[9] https://data.gov.rs/sr/terms/

More information about the Imports mailing list