A while ago I came across the natural=tree_group tag, but I can’t really tell how it differs from natural=wood, because the wiki doesn’t clearly say where the line between the two tags is. I’d like to understand the difference and then update the wiki accordingly.
What the wiki says
natural=tree_group: status in use, described as “Small group of trees”. It is meant for groups of trees in parks, urban areas, fields and meadows, or as landmarks, that are “generally not considered forests or woodlands”. However, it doesn’t say at what number of trees or what size a group becomes a wood.
The page lists cases where the tag should not be used. One of them is “Field copses bordering forests or standing in fields”, which should instead be tagged with natural=wood, landuse=forest, natural=tree or natural=tree_row. This contradicts the tag’s own definition, which names fields and meadows as typical places for tree_group.
Its own See also section describes natural=wood as “used for tree-covered areas, also for small groups of trees”.
The photo illustrating the tag looks to me more like a small natural=wood than a tree group, which in my opinion shows how unclear the difference is.
Questions for discussion
Do we need a separate tag for groups of trees that are not woodland? A typical case is trees with mown grass under their crowns in a housing estate or a park, where children can play, unlike an overgrown copse. Or is natural=wood (or individual natural=tree nodes) enough?
If we do, how should we tell them apart? By ground cover (grass vs. shrubs and forest floor)? By whether the crowns touch? Size or number of trees alone doesn’t seem like a good criterion to me. A small overgrown copse is still a wood, just a small one.
If we can agree on clearer criteria, I would update the wiki: add them to the tree_group page and mention tree_group on the natural=wood page.
If no agreement can be reached, I would at least add a note to the tree_group page that the distinction from natural=wood is unclear.
My view: tree_group makes sense for a group of trees over a lawn in built-up areas, often combined with landuse=grass. For copses and overgrown groups of trees in open landscape, I would stay with natural=wood. But I’m happy to hear other views.
Attempting to differentiate natural=wood[1] from natural=tree_group based on size seems unlikely to work well. I expect mappers would end up with varying ideas about the size that should qualify for each tag and data consumers would end up just treating natural=tree_group as another synonym for natural=wood and landuse=forest anyway. So we’d spend a lot of time and energy discussing/arguing for no benefit in the end.
I dug into this question last year in this topic: Poll: What makes an area woods/forest? Though the community appears split, it seems to be leaning more heavily towards not differentiating.
I’m very leery of recommending landuse=forest for any tree-covered areas. This kind of usage is fundamentally a misconception of what landuse actually is. Land use reflects some active human application and intervention in the landscape. An area where a bunch of trees just happen to be growing is not that. Even a bunch of intentionally planted trees isn’t necessarily a landuse, unless those trees are actively being used for something, and that something could well be something other than ‘forest’.
Traditionally, landuse=forest is strictly reserved for active forestry (and honestly ‘forestry’ should probably have been the tag to begin with). By contrast, a natural forest/wood isn’t a landuse at all. It’s an absence of use. Hence why the tag natural=wood exists. You could argue that non-forestry managed woodland counts as a type of landuse, but honestly how is a casual mapper going to know the difference between managed and unmanaged woodland? Better to avoid confusion and limit landuse=forest to strictly forestry.
and this is fundamentally a misconception of fact that keys in OpenStreetMap tags are not carrying strong meaning
natural=water is used for artificially setup water areas, we have amenity=prison which is kind of unusual case of amenity at best, historic=wayside_shrine is used also for modern shrines, historic=castle castle_type is used for classifying also palaces and manors…
This was the reason why Carto recently declined a request to render natural=tree_group (at least if mapped as a way - it would make more sense as a node).
It’s not often that myself (or anyone else) agrees with Carto resolution, so it’s prudent to acknowledge it. Introducing natural=tree_group was a non-solution to a non-problem.