Tree-filled riverbank: how to map?

I’ve been using "natural=tree_row" in this context for the last couple of years. Lines of trees randomly spaced along rivers, streams and railways are quite common where I map. The wiki is a bit confusing as the only picture examples are for “man made” lines of trees. I suspect that the “natural” version wasn’t thought about when the tag was developed. I did ask the community via a mail group at the time and the general response was to use natural=tree_row.

Initially, I only used a single way but if you do decide to use the above tag I would draw ways on either side which is what I now do as it fits better with other features. There is no requirement in the wiki for nodes to be where the trees are, it only suggest doing this if you want to map the individual trees which is quite problematic when they are randomly spaced. I guess an additional tag would enable you to distinguish between the random (natural) and the fixed width (man made) version, so people could render them differently.

Other practical issues I have found are distinguishing between a hedge and a tree row along streams. Quite a few hedges where I live haven’t been cut for many years so they can look like rows of trees. Looking at any historic maps available can be helpful in this respect. If there is a field boundary on the historic map then it is more likely to be a hedge. Occasionally I also find hedges that are in a poor state and look like a row of trees. Again, historic maps can help.

The best way to determine whether the feature is a wood or a line of trees to survey it. The decision can be a bit subjective but when the width of the random row of trees at the base is much more than about 5 meters I would probably map it as a wood.

Using “tree_lined” is probably easier but using a separate way enables you to connect other features (barriers etc) to the way. Something you wouldn’t do with the river way as it is supposed to represent the centre line of the river as far as I understand.

In short "natural=tree_row" and "tree_lined=*" are both likely to be acceptable, at this time.