Why ground effect, and the road doing half the work
Why proximity to a surface makes downforce far cheaper to generate.
Ground effect happens because the road itself becomes part of the shape squeezing the air, so a floor working close to the ground gets a share of its acceleration and low pressure essentially for free from the road surface underneath, rather than having to shape a curved surface entirely on its own the way a wing floating in open air would.
The road as the other wall of a channel
A wing generates its force by curving air's path with its own surfaces alone, one side of the wing doing all the work of speeding the air up and lowering its pressure. A floor working close to the ground shares that job with the road itself, because the gap between the underside of the car and the road forms a channel, and narrowing that channel, either by the floor's own shape or simply by the car sitting lower, speeds the air passing through it up and lowers its pressure in exactly the way a narrowing pipe would, without the floor needing to be anywhere near as aggressively curved as a wing generating the same force in open air. The closer the floor sits to the road, the narrower that channel becomes for a given floor shape, and the more of the acceleration work the road surface is effectively contributing on the floor's behalf, which is why the same underbody shape produces more downforce running low to the ground than it would running the same shape high up in clean air with no surface nearby to help it. A wing mounted well above a car's body, working purely in open air, has no such partner to lean on, which is one reason a wing generally needs a far more aggressive curve and angle to produce a comparable amount of force to what an underbody floor can generate with a comparatively gentle shape.
A playing card skimming the floor
A sheet of paper or a playing card dropped flat and allowed to flutter down to the floor often seems to glide and skim for an instant just above the ground rather than landing immediately, cushioned briefly by the thin layer of air it is squeezing between its own underside and the floor as it approaches. That brief skimming happens because the closing gap between card and floor is genuinely doing something to the air trapped in it, compressing and accelerating that thin layer in a way that would simply not happen to the same falling card out in open space far from any surface. A car's floor experiences a much more deliberate, controlled version of exactly this effect, the narrowing gap between floor and road actively shaping and speeding the air passing through it, with the road contributing every bit as much to that shaping as the floor's own surface does. Dropping the same card from further above the floor removes the effect almost entirely, since the gap never closes fast enough relative to the card's fall for the trapped air to build up any meaningful acceleration, exactly the way a floor raised too far from the road stops feeling any useful contribution from the surface beneath it.
Too close to the road, and the exit at the back
Ground effect explains why proximity to the road boosts the force a given floor shape can generate, but it does not by itself explain what happens if the car gets too close to the road, since at some point the channel between floor and ground becomes so narrow that the airflow can no longer pass through it smoothly at all, a separate problem covered by how the flow behaves once it is asked to squeeze through an opening that has become too tight for its own good. It also does not explain how that low-pressure air, once accelerated by the narrowing channel, is meant to slow back down and leave the car cleanly at the back, which is a question about the diffuser's own exit condition rather than about the ground effect that created the low pressure in the first place. Ground effect and the diffuser's exit are really two halves of the same story, one explaining how the low pressure gets created cheaply and the other explaining how it is safely handed back to the surrounding air, and a floor only performs well when both halves are working together.
Ride height, test rigs and rolling roads
Ride height is treated as one of the single most sensitive setup variables on a car that relies on underbody downforce, since even a small change in how close the floor sits to the road changes how narrow the effective channel is and therefore how much of the road's contribution the floor is actually receiving. It is also why a floor's performance measured on a rig or in simulation far from any ground plane tells a team very little about how that same floor will behave once mounted on the real car and lowered to its intended running height, because so much of the useful force is coming from an interaction with the road that simply is not present without a real, close surface underneath it. Teams testing a floor design therefore often build a simple rolling road or a fixed ground plane into their testing setup well before they have access to anything more elaborate, since a result gathered without a proper ground reference underneath the model can point the design in the wrong direction altogether.