How to map trees in open street map?

I am confused a little bit with: how to map a couple of tree nodes. I am using either iD or JOSM.
According to the wiki tree page, the node should represents the location of the tree’s trunk.
What confuses me is that with the underlying satellite imagery we can only see the tops - crowns of the trees.
So how can one identify the location of the tree’s trunk? It it can not, does that mean that all tree nodes in openstreetmap are positioned somewhere in the middle of the crown from the satellite imagery?

You also could use winter satellite images to identify the location of the tree’s trunk.

It is very reasonable assumption

This isn’t much different from buildings, where the roofs also tend to be offset in imagery and you need to correct for that when drawing building outlines. Of course, buildings are a bit easier because you often see parts of the walls in the image, giving you a hint how much you have to offset the outline.

For trees, you can derive it from the location of other features (e.g. you know that the tree is in the middle between two footpaths). You can also just get a feeling for the amount of distance you need to subtract for a tree of a given size. Or just go out and have a look at it. Satellite imagery isn’t the only possible data source for OSM, after all.

I’m sure its a common beginner’s mistake and therefore quite prevalent in the data. But once you are aware of the issue, I feel it should be possible to achieve good enough precision. It doesn’t have to be perfect.

Thank you for the quick and detailed reply wowik!!

Does JOSM have the winter satellite images? If yes then which one of these is the winter satellite image:

Tordanik, thank you for the detailed reply!

I didn’t understand these two sentences:

Amount of distance to be subtracted from what?

I didn’t understand you this part. Do I need to somehow record the GPS coordinate just next to the tree trunk, then somehow use that coordinate to add a tree node?

Imageries consist from many images which obtain in different seasons. So in some places we can found winter or autumn images in some imagery. Just to try all

You are, again, asking questions that no-one else would bother to ask.

There also seems to be an assumption that the normal way of mapping with OSM is remotely, from aerial photographs (real satellite images generally are not good enough for individual trees). The preferred method of mapping is to actually visit the place, and make measurements. That way, you can account for sloping trunks, as well.

By subtracting, they meant take the estimated position error due to the obliqueness of the image (parallax error), and use it to correct the position of the root.

In terms of visiting the place, any of the techniques used by professional surveyors (I doubt many of them rely on aerial imagery, even now) can be used, including methods that pre-dated GPS. In practice, very few people are going to care about small errors in tree trunk positions, anyway. If you cannot measure the error on the ground, it probably doesn’t matter.

Thank you wowik. I didn’t know that.

Also thank you for the detailed explanation hadw.

Is mapping by actual visit of the place, preferred mapping method for any sort of OSM object/OSM element? Not only tree nodes?

Is mapping by actual visit of the place preferred mapping method even in cases when one needs to map a building ground outline, and does not have an access to GPS?

Personally, at the resolution at which OSM works, I think too much effort has gone into mapping building outlines, probably because they can be armchair mapped. I don’t think that the outlines that can be recovered from aerial images are accurate enough for any serious use of the map, and, for example house numbers would be much more useful for routing, etc. At best they give a general impression of the building and its relation to other buildings.

Serious users of building outlines need a level of details that really needs a theodelite and tape measure/laser range finder,and probably also benefits from having access to the architect’s drawings. You cannot really get an accurate outline for the small block of flats in which I live, from aerial imagery, and the shopping centre in the local town requires quite detailed examination on the ground to get it anything close to correct.

Finally. the buildings for which I would really want an outline now are too recent to appear on any aerial imagery.

It’s worth noting that for a large amount of trees, sometimes mapping them individually is not needed, and people wrap them all with a landuse=forest or natural=wood area.

Each technique has advantages and disadvantages. Some people within the community seem to really dislike armchair mapping, others find it acceptable.

Aerial mapping allows you to map a massive amount of buildings and streets within a short time.

An actual visit gives you addresses, names, amenities and shops, and other small things you can’t see from Aerial. The information is often more up to date too. But mapping lots of buildings outlines or streets can be a lot easier and can take an order of magnitude less time with Aerial.

See this and this for more details about mapping techniques.

Thank you once again for the clarification hadw!

SwiftFast thank you too for the very nice reply!
I also hasn’t been aware of those two articles. I will definitely have to read them.

Talking of armchair mapping and trees… a few months ago I was playing with a tree finding algorithm in R which can identify trees in LiDAR data and output a point set of their locations. There is also an algorithm I have seen in research literature which attempts to classify trees as coniferous or deciduous from crown shape using point cloud data. I can programme but, with the time I have available for such experiments, my maths and knowledge of R isn’t yet strong enough to turn what I have found into usable tools.

If OP or anyone else who should happen upon this post is interested in that sort of technique please let me know if you want to collaborate! Trees are nice features to have on a map because in some areas they serve as landmarks due to their height.

Hi Peter,
I know nothing about the R language, nor I could help with the algorithm you mentioned but of these two sounds very interesting:

Can you point me to that tree finding algorithm and publication about identifying coniferous or deciduous trees?

Thank you.

So the R code was from the rLiDAR R package:

I cannot remember the exact article for the second part but try this:

And check out Google Scholar:

Thank you Peter!