The biggest issue with multi-tile trees as I see it is coming up with a good and intuitive way to represent them in ascii that doesn't completely fry the brain of any new player when they first take a look at them. I don't even play DF without tilesets though so not sure if I have anything to add there though.
It all also depends a huge amount on how far Toady wants to stretch towards realism as far as trees and forests go. My knowledge on this comes from me studying biology, ecology to be specific. Basically, I propose 2 different growth patterns for most trees depending on if it's in a forest/jungle biome or a more open one such as a grassland. Trees that grow in the open have free access to sunlight, and thus don't have to put much energy into growing taller, but instead grow more in width and spread its branches to the sides to reach in order to maximize the sunlight it can get.
Example of free growing Oak tree:
Example of free growing Pine tree:
A tree growing in a forest on the other hand has to compete over the available light with the other nearby trees, and as such put more energy into growing taller (trunk and branches being much thinner as a result), trying to reach above its competitors to snatch as much of the available sunlight as possible.
Example of forest growing Oak tree:
Example of free growing Pine tree (the trees around having been felled):
Anyhow, the point I'm trying to make is that I think these 2 classes should be represented in the design of multi-tile trees. Reality is ofc not quite as clear as this and one can find trees in all shapes on the scale between these extremes, but this should be a sufficient approximation for DF.
As for the actual trees, I assume a tile being 2*2*3 (?) meters as stated by Toady for current physics calculations. They'd all start with the base trunk ofc, which really shouldn't ever need to be more than 1 tile other than for "giant" trees for elven forest retreats. It'd also be suitable to not have any of the canopy for the first z-level for most trees to keep things relatively clean). In a forest you'd then have x amount of z-levels of mostly trunk up to the actual canopy which would spread 1 tile out in each direction ish in a thick forest, creating a relatively compact forest canopy. What tiles to use for this I can't really say, any ideas? A freegrowing tree on the other hand would have its canopy start on the second z-level, spreading 2 tiles out in all directions (save some variation to keep it from being too uniform) and span maybe 2-3 z-levels up depending on the tree and its age. Possibly using some other tiles to represent thicker branches, also indicating you could easily climb these whenever climbing becomes a factor. Also, the current tree tiles could be used to indicate younger trees, ie the ones not yet spanning more than 1 z-lvl but that' still give some minor amount of wood if cut.
Speaking of woodcutting, I propose that when you fell a tree, a direction without any constructiones etc obstructing would be "chosen" by the dwarves cutting it, them trying to make sure it falls in that direction, and making them take care not to step into that area as its falling. Of course, unskilled woodcutters could fail with this, leading to Fun to be had. As it hits the ground, a small cave-inesque effect would occur. On the ground you'd now have a tree object of some sort of suitable length, which the woodcutters could then chop up on the spot into smaller more manageable logs to carry into the fort (lenght and width factoring into how much wood you get out of it). These changes would probably lead to getting much more wood out of 1 tree, which I however consider a good thing. Tree growth should be slowed to more realistic figures and this would mostly just balance itself out. It'd also prove a more interesting challenge to clearcut a forest with improved felling mechanics. Overall you'd get the same amount of wood for the time spent (unless tweaking is needed), just that there'd be more work involved per tree.