Best way to map a national park/AONB boundary

I’m in the process of adding the boundary of the Howardian Hills Area of Outstanding Natural Beauty with data provided by the AONB Office themselves. I’ve done about a quarter of it so far, I would say.

My question is, am I doing it right? It’s hard work and I don’t want to have to redo it!

Where possible, I am following existing map features such as roads and streams and adding them to a relation ( Where there is no particular feature to follow, I’m creating an untagged “linking” way in the same way as this one here in Snowdonia ( so that I get a continuous line and ultimately an unbroken boundary.

In a previous posting, ( I’d broken the rendering of a wood by snipping its boundary into a segment that I could add to my relation, and in a reply, Nightdive suggested that I simply create a “linking” way round the wood.

What does everyone think? Which is the best way?


As the ‘author’ of the Snowdonia boundary, I have to agree that this is the way to do it! The main purpose of these linking ways is to close off the polygon so that it appears on the various renders. Your relation looks fine to me, but I would add designation=AONB so that in the future we can potentially render national parks and AONBs differently.

Snipping ways and polygons and replacing them with relations is widespread practice: take a look at a roundabout with a range of bus-routes operating over it.

We have a long way to go in adding National Parks, AONB, and National Scenic Areas. I think the only way is to get an approximation of the boundaries into OSM, and then use maps to ask people who are likely to know to refine these boundaries more accurately.

One last point many sources of boundary data will be tainted as being derived from OSGB data, but knowledge derived and communicated by people is not. This may complicate how you acquire information which from the AONB office.


Splitting existing areas (area= the first and the last node are the same) and using existing ways for a boundary is bad because such boundary lines are very often damaged by other mappers.
It might not be a big problem for areas with only a few mappers but this is a problem in big cities in germany. There are suburb boundary lines that are on street ways and they are damaged very often with merging ways or if a street get changed.
Other boundarys have their own independent ways and they get only damaged if someone works on boundarys.

FYI: I repair damaged boundays nearly every week in germany ( ) .
Splitting Areas and putting them together with a multipolygon-relation makes is worse.
It’s very difficult to understand this areas. You have to load all the members of the multipolygon area and highlight them to understand the area.
ok, this isn’t a bug problem for a single forest but look at all the areas here :
A roundabout is not an area (AFAIK) and putting busroutes on the streets is of course the right thing to do.

I can not see a benefit from splitting existing ways to add them in a boundary.

It’s worth noting that often the boundary is not actually the centerline of the road, but the property line on one side.

Academically you are correct, but in relation to OSM this comment worries me a little.

In my experience locally much of the OMS mapping (in the UK) is derived from copying from old OS maps at a low level of zoom and can be 10s of meters out. Even when I survey my mapping with an accurate GPS using WAAS/EGNOS and a clear sky signal my accuracy still likely to be 4/5 meters out, therefore I would not be able to map a road sufficiently well to be able to distinguish right/centre/left of it.

Are you therefore implying that I shouldn’t be mapping OSM?

I’ve found that NPE isn’t always too well lined up but I don’t know whether this is inaccuracies with the maps or with the scanning. The more recent ones (25K and OS7) line up beautifully. By always uploading all your traces (rather than just using them offline in JOSM, say) it is easy to see an average and judge for yourself how well they line up. You can move the background map in Potlatch by holding down the spacebar and dragging it.

The line of nodes represents the centre point of the way but I would trust your satnav a bit more! With many traces up and down a road you can often see the lanes and the way should follow the average.

This is one of my best - you can see where I walked up one side and back down the other. The ways themselves, obviously, run up the middle.

No! But use all the information to hand - the gps traces, OS maps and aerial photos - and do the best you can.


That is a very good point.

Back on topic, there would seem to be 2 styles:

  1. Use existing ways (roads, rivers, footpaths etc) and add them to a relation. Where there are breaks then use “linking” ways to join them up to give a completely enclosed area. This has been used for Snowdonia, the Brecon Beacons and the Lake District amongst others.

  2. Create a separate way running close to bounding features but distinct from them. This has been used for Dartmoor and is suggested also by Nightdive.

Number 1 seems the most common, but I couldn’t find an example of the problem I had with the wood. I am getting the same problem with the boundary running round the outside of a village at the edge of the residential land use. The edge of the land use is hard to estimate from the OS maps and is obviously out of date anyway.

Snipping an existing boundary and then creating a multipolygon to repair it is a bit of a pain and is not intuitive. I’m also conscious of future generations of mappers not having a clue why this has been done. A separate way is less elegant but more intuitive to maintain.

I’m tempted to use #1 for the majority and then create ways round existing areas that would otherwise need to be snipped, but I will discuss it with the AONB office. If the wood were to change, or more likely the edge of the residential area grow, would the boundary move with it? Don’t know.

This is a good point, and probably something I’ve neglected to document the reason for my initial approach.

#1 is to be preferred for not only the reasons advocated, but because often features on which a boundary is based move, but the boundary does not (rivers are the classic example). However, creating additional ways which directly overlay or in close proximity to other ways makes them very hard to edit. Furthermore many of the boundaries being created with this technique are extremely approximate (National Park boundaries in England and Wales may be several kilometres from their true positions), because of the general absence of non-OSGB derived data.

#2 has therefore been used as a quick way of drafting boundaries without encumbering the data with lots of additional ways which may be removed as time progresses.

Dartmoor was, I think, originally mapped as one huge way. With relations, the relation can be retained, and its members gradually all made specific boundary ways as the actual information is refined.

Generally it’s inaccuracy with the maps. The NPE images in Potlatch have been rectified to the National Grid every 5km so should be pretty faithful to the original.

Well I’ve finished the boundary at long last. The relation is 383665 and I’ve checked it’s complete at There were some tiny gaps here and there where roads appeared joined when they weren’t etc.

There’s not much sign of it rendering yet though, even though it’s been over 24hrs. Normally I find Mapnik starts within a couple of minutes, but do these things take longer? I think I’ve tagged it right.

Any ideas?


The first and last ways were not actually joined, so it was continuous but didn’t form a loop. Now it works; I’m not sure if leisure=park should be added to the relation (this colors the inside light green in Mapnik).

Thanks NE2, well spotted! The one final little bug.

The tagging of national_park renders it green on Mapnik, you can see it there now :slight_smile:

Did I marry a Mr. Vclaw without knowing? :slight_smile:

oops, sorry!

15hr day yesterday and 13 today!