[Talk-GB] Administrative boundaries (from Re: traffic island mapping / harmful detail?)

Colin Smale colin.smale at xs4all.nl
Tue Apr 6 08:18:48 UTC 2021


>     On 04/06/2021 12:51 AM Tom Crocker <tomcrockermail at gmail.com> wrote:
> 
> 
>     Thanks Colin. 
> 
>     On Mon, 5 Apr 2021, 22:44 Colin Smale, <colin.smale at xs4all.nl mailto:colin.smale at xs4all.nl > wrote:
> 
>         > > 
> >             > > >             ... On the other hand the admin boundaries provided by Ordnance Survey are surveyed with extremely expensive kit with an accuracy of around a centimetre. These coordinates are put directly into OSM, without any tracing or surveying by mappers. If you see that a boundary runs parallel to a road's centre line but not on it, there are several possible explanations: It may be that the aerial imagery is correctly located and the boundary is indeed where you see it, but it is seems more likely to me that the imagery is offset by a bit and that in fact we should move the highway to match the boundary, even if that looks wrong when referenced to the aerials. 
> > > 
> > >         > > 
> >     >     In this case it was a stream that wiggles more than the legal boundary. OS opendata streetview is also differently aligned to the boundary. Anyway, I'm happy for the stream and boundary to be different.
> 
Wiggles remind me of another point: data generalisation (filtering out "unneccessary" detail by reducing the number of individual nodes). As OSM, in common with most GIS systems I know of, only knows about straight lines between points, only the nodes have any value in terms of location. Between the nodes, we draw a straight line, but that does not imply that the object concerned actually follows a straight line between the two points. It is tempting to assume that a version of a stream with more points is "better" than a version with less points, but that may depend on your interpretation of "better".

> 
>         > > 
> >             > > >             I also agree it would be really useful if the boundary could be in some separate layer (even if, in principle, editable by all). I was recently ungluing some landuse that had been stuck to the boundary and realised at the end I had carried the old nodes with the landuse rather than leaving them on the boundary and had to figure out how to swap them back.
> > > 
> > >         > >         Landuse boundaries don't correlate that well with admin boundaries, which happily cut across fields and towns. Even at field boundaries you cannot see if the admin boundary follows the centre line of a centuries-old hedge or the ditch behind it for example. Putting the admin boundary to one side, how well can you determine the actual landuse boundary? Do you want the hedge to have a distinct land use value such as field_margin or will you include it in the field?
> > 
> >     > 
>     I think we've got our wires crossed here. I'm agreeing with the suggestion of others that having the boundaries in a separate dataset, not immediately editable along with the rest of the map, would be a good idea, but I appreciate it may be entirely infeasible.
> 
Sorry if I went off on a tangent, I do that sometimes...

I would also love to see some layering in OSM. I'm speculating here, but I think many people like to share nodes and ways between objects of different types because they think it's "efficient" in some way, maybe in terms of data storage. Sometimes it is appropriate because the association is by design, such as a shared node where roads meet at a junction. The endpoints of the roads are not independent in real life, you cannot move one without moving the others. However a boundary line and a waterway or highway are not inextricably linked by design. They may have been once, but administratively they now live independent lives. If a road is re-routed or a junction is remodelled, that in itself does not move any boundary. It is possible that the boundary gets moved to suit, but that is an independent change, which may happen long before, or long after, the civil works, or it might not happen at all. The same with rivers, streams etc - which besides being rerouted by human intervention, may also move by themselves.

Putting (for example) boundaries, waterways and highways into different layers can help to reduce inadvertent (or ill-informed) changes to different object types. But I think this kind of layering could be implemented by editors placing object types into virtual layers, allowing only one layer at a time to be selected, and automatically unjoining nodes where a node that is shared between layers is moved, and being reluctant to share nodes between virtual layers. From a technical viewpoint this doesn't sound like rocket science although I recognise it will be quite a bit of work for the editor maintainers. I am more afraid of the discussion about the definition of the layers and what objects go in what layers....


>     In this case a forest crossed a boundary and had been glued to it where it crossed, accidentally or otherwise. I was ungluing it to realign its edge to the land registry data which shows where the wall runs. 
> 
>     Since you ask I would generally glue fences, walls or hedges to landuse  boundaries such as field edges. I only leave a gap if a track or path is fenced or walled out of the field. When I'm in micromapping mood I think there's a question about where to stop the field when there's a hedge next to a wall but would personally put the hedge inside the field and form the boundary along the wall. Sometimes a wall is no longer a field boundary because there's a new fence. Anyway, I'm happy for others to do what they do on this.
> 
>         > > 
> > 
> >             > > > 
> > >             Out of interest, how often, if at all, are the UK boundaries redrawn in OSM to undo accidental or otherwise 'improvements'?
> > > 
> > > 
> > >         > >         Not regularly, but I can say this takes up quite a bit of my time.... I monitor all changes to admin boundaries in the UK and judge whether to fix "minor improvements" or not. I try to get actual (legal) changes implemented as soon as possible after the boundary data becomes available and I also have a backlog of things that need "fixing"...
> > 
> >     > 
>     Thank you very much for doing this 👍
> 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20210406/3e499cfb/attachment-0001.htm>


More information about the Talk-GB mailing list