I have pretty rocky ground and digging roots out with a (tractor) backhoe can be pretty involved.
I can trivially pull over/pull out pine trees, but these guys don't behave nearly the same (besides the fact that they'll break off if you pull from a high enough point that leverage from the height is useful).
Some have noted that a wheel just changes direction of pull - correct, but still helpful since trees have apparently not considered that there's more to life than avoiding being pushed over and someone may try to pull them upwards... I wonder if trees in parts of Africa are different, since they have to deal with Elephants which sometimes pull trees out. A chain pulled over a wheel could break with an upward velocity component vector if the break is between the tree and the top of the wheel, while a break between the top of the wheel and the tractor should have the chain portion only moving towards the tow point (though there's a potential for rebound off of the ground, but this seems unlikely for a heavy chain in the duff).
The I-beam (that could be modified to an upside down "t") can be set up at an angle to provide significant mechanical advantage (when placed at an angle to the ground greater than 45°, there will be advantage; the higher the angle the higher the advantage, though for less distance). I have a heavy metal pipe A-frame thing that I thought of yesterday that would probably serve this purpose very well also.
I'm thinking of getting some grade 70 chain for the tree-to-leverage-device portion, since the tree-to-device portion of chain will be under much higher tension, and use my 3/8" grade 43 to pull on the beam/A-frame.
A break in the tree-to-device chain could have upward travel but that'll be localized. A break in the pulling part would be similar to a break in the similar part of the wheel setup