The atmosphere is shielded from UA in the same way that people on the surface are. UA posits that less-massive objects (including plants, the air, airplanes, etc) are "shielded" by more massive objects from the spontaneous acceleration force.
Applying the equivalence principle, UA says that objects near a massive object like the Earth experience a force in one particular direction, "downward" on Earth, unlike gravity, which pulls you toward the mass itself. As you get farther from the Earth's surface, the magnitude of this invisible force decreases. I recognize that this is not the UA presented on the wiki; however, before you accuse me of ignorance and/or misrepresentation, this model makes exactly the same predictions and assumptions as UA, per the equivalence principle. Essentially, FE proposes in UA a form of gravity that is normal to the flat surface of the Earth and constant that dies with increasing altitude without explaining its origins. Of course, it cannot be due to the interaction between masses, because they deny that (it would pull the Earth into a sphere after long enough).
It's wrong to talk about axes in isotropic space. Of course in FE, there is a preferred axis, the direction of UA, or the direction opposite it. Then the remaining two axes must simply lie in the orthogonal complement of the space spanned by the basis vector in that direction, but again the choices of directions of the two would be arbitrary, so long as they spanned the requisite space. Or you could just dispense with the remaining two axes, and use a curvilinear 2D coordinate system. But there would be a mathematical preference against a 3D curvilinear coordinate system, precisely because space wouldn't be isotropic under FE, and one straight axis would be physically defined. Essentially, UA would imply the existence of a universal compass. This is quite an assertion; there are various problems that could be elaborated on if anyone made a thread in FE Debate.
Of course, in RE, there is no "z-axis" or "y-axis" other than what you define them to be (I could point the z-axis toward the Sun or toward Tom Bishop's house, and both would be equally valid). The point in Galilean motion and special relativity is that all coordinate systems are equivalent, and general relativity generalizes this by describing the curvature of space using the stress-energy tensor (again absent coordinates).
The variation of g is taken by FE to be the result of "celestial gravitation" -- A standard RE rebuttal is that in essence, they have an invisible force that's just as inexplicable as gravity itself. So much for being simpler. Of course that doesn't explain why it varies almost exactly according to the amount of centripetal acceleration on a rotating sphere at 1.15 x 10^-5 Hz. No FE has written out a formula for "celestial gravitation" that describes the amount of force between two point particles. This is a complete dealbreaker for me, because it doesn't make specific predictions.