Hello,

I did some mapping in Wallis last week, and obviously noticed the landuse=forest mapping.
I must first say that I am very impressed with the size of the area mapped with better than normal accuracy.
But I have to put (very) cold water on this enthusiasm.
The forest mapping in Wallis is not accurate; it is not micro-mapping; it is a fancy sine wave around the actual (usually straight) land use; over 95% of the nodes are based on shades, non-orthogonal projection of the canopy on the ground, or pure invention. They are not based on terrain, and removing most of them actually improves the accuracy of the map.

Here is an example. If we look at this (non-free) data:
https://www.google.fr/maps/@46.0985256,7.0849358,3a,75y,103.54h,95.62t/data=!3m6!1e1!3m4!1s3ZO7vmctMaMeZX1HLsG7qQ!2e0!7i13312!8i6656
a straight line of trees and a straight fence marking the boundary between the forest and road land use are shown.
The most accurate mapping would have been to use the fence as the boundary of the landuse=forest, but instead there is a sinusoidal curve that even has a few nodes on the motorway asphalt. It is very noticeable with the swisstopo SWISSIMAGE now that most of the trees have been cut down.

Marc