I recently received an e-mail from a fellow mapper complaining that I had deleted a landuse=farmland he had created. When I reviewed the area, I had deleted the field so that I could add a wood and other features without needing a very significant session to split up farmland areas, detach nodes and ultimatly need multipolygons. The area in question is north east of Biggin Hill. and has each field separatly defined as landuse = farmland. The mapping principle I use is that land in the UK is by default farmland unless tagged otherwise. I only use landuse=farmland where the landuse is exceptional such as within an urban area. I split fields using barrier=hedge, waterway=drain or barrier=fence as appropriate. See Holderness in East Yorkshire:

I would appreciate comments and views on the communities preference:

  1. Use landuse=farmland for each field.
  2. Use landuse=farmland only where the farmland is the exception and define field boundaries with the approriate physical feature.


There is no default value for landuse=*. Therefore, every occurrence of farmland should be tagged as landuse=farmland, regardless whether it is exceptional in an urban area or a large rural area.

I think you mis understood the question. To explain the issue more fully:

England landuse (as measured by the office of national statistics) is:
87.47% Farmland and forest.
1.14% Domestic Housing
4.27% Domestic Garden
0.66% Non domestic buildings
2.23% Roads
0.11% Paths
0.14% Railways

The map would be more accurate and use less data if the outline of the UK became a multipolygon with the tag landuse=farmland. Every area defined within it then need a inner polygon though. The alternative is to create many connected polygons of landuse=farmland. The extent however would be so great that most other polygons would be connected. The result would be very difficult to edit, for example: some polygons would not be selectable, node edits/additions would need to be disconnected, duplicated, repositioned and re-joined.

I don’t think that such a huge multipolygon covering the whole UK would be maintainable. Everybody who edits some landuse within the UK would have to take care about this multipolygon, and it is very likely that someone breaks it by mistake. The standard way is to use smaller landuse areas, because they are easier to handle.

However, much it makes your work easier deleting others objects to do so is not the best way to go about it: particularly if a user has spent a lot of time and effort as in this case.

You may feel that farms do not need to be explicitly mapped, but others are interested in such things and the information which can be derived from them. I think it’s important to respect others mapping peccadilloes unless they cannot be verified on the ground.

Personally I think farms should not be mapped field by field: the barrier=* tags are much more useful to achieve this type of result: and are much more informative, whether for walkers, conservationists, or just general cartography.

A default value for farmland is not really very sensible. Are grouse moors or upland sheep pasturage really farmland? This also includes large areas dedicated to the military, nature conservation, wetlands etc. All of which we map in different ways in OSM (and this does include the South East see the Surrey heaths).

You could have probably plonked your woods down without worrying about the farmland polys: I’m pretty sure that mapnik renders farmland at the bottom of the ‘painter’s algorithm’ stack. This is certainly the approach I took with extracting landuse data from OSM (see for example). Of course it is better to do it properly, but when time constrained adding new data can be done without deleting existing stuff.

Using barrier=* tags seems like a nice option. There is a lot of farmland that needs to be mapped in my area and I’d like to use that. One question: When a barrier crosses a path or track, would this result in a blockage for autorouting? In that case, I’d need to be very careful that I don’t block anything when I have farmland fences on both sides of a road.

Where there’s a gap in a hedge or a fence map the gap (i.e. don’t put hedge or fence where there isn’t any in reality) and it won’t block any road or track.

Some routers (from memory, in some circumstances Strava’s may be one) do struggle with barriers such as gates, but if that happens that’s a bug in the router.

Thanks. Perhaps using barriers is a risky option then, as roads may accidentally get blocked when they are later moved over a barrier. Would a road still be blocked if a barrier touches the road from one side (i.e. barrier and road share a node), but the barrier does not continue on the other side of the road?

If a road shares a node with a way that is a barrier, what happens depends on the particular router, of which there are many. My guess is that it won’t affect routing.

However a barrier (such as a fence) and a road sharing a node is a bit odd - it sort of implies that the fence starts in the middle of the road, rather than to one side of it.