Ground-effect analysis is generally completed using the method of images, where we imagine an opposite geometry providing a flow that interacts with the original geometry.
The 2D case
Katz and Plotkin have a good explanation; here is their diagram for the 2D case:
For a flat-plate airfoil, the math works out to
\begin{equation}
C_\mathrm{l} =\frac{\pi \alpha \left(c^2-8 \alpha c h+16 h^2\right)}{2 h (4 h-\alpha c)}
\end{equation}
where $c$ is the chord, $\alpha$ is the angle of attack, and $h$ is the height above the ground.
The 3D case
In the 3D case, we model the entire wing (or aircraft) as mirrored. The math here gets quite a bit more complicated, but I can summarize Phillips and Hunsaker, who provide the following figure and equation:
For an untapered flat wing with small $\alpha$, the lift coefficient in ground effect $C_\mathrm{L_G}$ is about
\begin{equation}
C_\mathrm{L_G}\approx\left(\left(\frac{261.562}{\left(\frac{b}{c}\right)^{0.882}}-\frac{1.944}{\left(\frac{b}{c}\right)^{0.165}}\right) \left(\frac{h}{b}\right)^{0.787} e^{-9.14 \left(\frac{h}{b}\right)^{0.327}}+1\right)C_\mathrm{L}
\end{equation}
where $C_\mathrm{L}$ is the lift coefficient in the absence of ground effect, $h$ is the height above the ground, $c$ is the chord, and $b$ is the wingspan.
Comparing the two
I plotted the 2D equation for an airfoil with $c=1$ along with the 3D equation with $C_\mathrm{l}=C_\mathrm{L}$ for a wing with $c=1$ and $b=10$ (for any reasonable aspect ratio, this trend is similar):
As expected, far away from the ground and very close to it the 2D and 3D cases converge. The 3D case gets a bit more help from the ground in the intermediate range. Because we have only considered inviscid flows in the mathematical analysis, we know that the induced drag is the major difference between the 2D and 3D cases.
