[Talk-se] Mapbox Organized Editing

Thu Mar 20 20:56:51 UTC 2025

While there quality in NVDB does vary (it's a collection of multiple 
sources) and geometry simplifications does occur, it's on average very 
good. You cannot get the same accuracy from Bing or ESRI imagery, if you 
want to get similar precision from ortophoto you would need 
lantmäteriet's own. Sure if you have your own diff-GPS system you can 
get more precise, but most maps from some sort of orthophoto, usually 
with limited elevation correction leading to fairly large positional 
errors.

Ideally you should use NVDB + Orthophoto + strava heatmap for sanity 
checking + local knowledge where available. I've mapped tens of 
thousands of roads in this manner in northern Sweden and quality of NVDB 
is very good, while roads mapped on (especially older) ortophoto in 
hilly regions can be very far off the mark. In city geometry stuff get a 
bit more complicated, because there may be simplifications made for 
navigation which may not match with OSM tradition, and there you need to 
deviate at times. We have an ongoing manual import project with 
documentation of limitations and caveats. You should never import 
blindly, that's why the NVDB import project (or rather manual merging) 
takes such long time. If you see errors like duplicated geometries, 
clearly the mapper hasn't paid much attention to his or her merging 
effort. Any source if misused leads to poor results.

It's best to use multiple sources and if working in Sweden NVDB should 
for sure be one of them, and if the only positional data you have is 
NVDB and an ortophoto from Bing or Esri, NVDB is nearly always better 
positioned, with the exception of geometry simplification possibly 
occurring on multi-lane roads, or sometimes very new forestry roads with 
preliminary geometry and other rare exceptions. Naming of forestry roads 
can be wonky in places, as the reporting can be patchy and approximate 
depending on forestry company responsible for the reporting. Naming of 
roads in cities are usually of very high quality.

The problem with OSM in Sweden is that the coverage can be patchy in 
places where there are few active mappers, and some mapping can be 
rather outdated. The big cities are usually well covered, or home cities 
of super mappers. There are large areas in Sweden which is only 
occasionally paid attention to, and there mapping work from people 
without access to the ground is needed, and in such a situation using 
NVDB in a responsible way as one of the key sources is the best way to 
go.

However, if the mapper doesn't really want to make the effort to use 
NVDB responsibly and not knowing about its limitations or how you relate 
to existing mapping, it's best to not use it of course.

/Anders

On 2025-03-20 13:07, Peter Svensson wrote:

> I disagree. The only things that shows exact locations of Swedish roads 
> are the ground itself. Everything else is approximations and 
> simplifications which are by definition "not exact". I've lost count of 
> how many things I've cleaned up after various NVDB import efforts, such 
> as incorrect names and duplicated/questionable geometries.
> 
> I encourage use of other sources than NVDB for mapping missing roads.
> 
> With regards,
> 
> zvenzzon
> 
> Den tors 20 mars 2025 12:31Markku Siipola via Talk-se 
> <talk-se at openstreetmap.org> skrev:
> 
>> Your answers shows you are not aware of NDBD, the National road 
>> database. This source should always be used, and not any source. The 
>> NVDB shows the exact locations of Swedish roads. It's waste of time of 
>> not using this db.
> 
> Den 2025-03-20 kl. 08:38, skrev Valeria Yukovich via Talk-se:
> Hello, Snusmumriken! Thank you for the questions.
> We will be using available sources which are appropriate for mapping on 
> OSM (satellite and street-level imagery, for example). Besides that, we 
> will be also using Mapbox telemetry data and GPS traces in cases when 
> there is limited evidence. More information about Mapbox telemetry data 
> can be found here: https://www.mapbox.com/telemetry [1].
> Speaking about our experience with other countries, we mostly 
> encountered missing roads in newly built residential areas and around 
> commercial objects, which we did not map in cases if the evidence 
> (street-level imagery or up-to-date satellites, information from the 
> telemetry) was not enough for mapping.
> 
> On Wed, Mar 19, 2025 at 12:18 PM Snusmumriken 
> <snusmumriken.mapper at runbox.com> wrote: On Tue, 2025-03-18 at 16:46 
> +0300, Valeria Yukovich via Talk-se wrote:
>> Hello from the Mapbox team!
>> 
>> Our team has been mapping missing roads since May (as described
>> here: https://community.openstreetmap.org/t/mapbox-organized-editing-
>> missing-roads/113678) and currently we are preparing to expand the
>> effort to Sweden.
> 
> Hi
> 
> I have basically two questions:
> - What sources will you be using in Sweden?
> - What are the most common problems you've encountered in other
> countries?
> 
> _______________________________________________
> Talk-se mailing list
> Talk-se at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-se
  _______________________________________________
Talk-se mailing list
Talk-se at openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-se
_______________________________________________
Talk-se mailing list
Talk-se at openstreetmap.org
https://lists.openstreetmap.org/listinfo/talk-se

Links:
------
[1] https://www.mapbox.com/telemetry
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-se/attachments/20250320/edd0fb55/attachment-0001.htm>