AFAIK 3D modelling of the gas is way too complex for current computing power and the initial conditions are way too sparsely known.
Consequently all wx models use simplifications, developed over many decades, which attempt to represent the main behaviours. So e.g. the models give you
This is what GFS gives you and AFAIK every wx model is the same.
The original GRAMET (called that because it was done at the Univ of Granada) plotted a graphical interpretation of the four (or whatever) cloud types. Actually Meteoblue was even earlier (no idea which URL that lives on; they tried to moneytise it) and that also used GFS. But all this is largely guesswork; you don’t actually get a 3D data set showing where e.g. IMC is. The programmer creates icons for “vertical sausages” of various sizes and drops these onto the chart to depict convective cloud, etc. Different “gramets” use different equations, hence give different results.
GFS has something like 64 vertical layers each about 3000ft think istr.
So it is 3D.
This isn’t evident in the plotted data, however. It still shows the “CGI” shapes.
Most current models today have lots of layers. Only, not a lot of applications using the model data are capable of displaying them all or construct a 3D visual model out of them.
Yes, weather models are 2.5D at best compared to 3D computational fluid mechanics models, tough the inputs for initial conditions and outputs for visualization are obviously 3D
In global forcasters like GFS, there are no diffusion terms along the z-axis but there are linear conviction terms and this is done on purpose as a proper 3D model will have too much vertical mouvement or vertical pressure glitches versus pressure stratification or pressure vertical profiles that are observed in reality (the z-axis is parametrized in linear decreasing pressure coordinate instead of distances and you only need to know the pressure on the surface)
