I spent some of my driving time thinking about modeling these things recently.
I don't think you would have to solve fluid dynamics (which is what the current fluid system does, in a piecewise, not-really-right manner) for air pressure and wind to work. Air pressure normalizes at the speed of sound, which has got to mean effectively doing the entire map instantly -- that is to say, there's no need to model sonic booms, explosive shock waves and such -- in an air pressure sim. So 'all' you need to do is to be able to set realistic boundary conditions along the edge of the map and solve the steady-state situation. To do the steady-state situation, and to have it be interesting, though, you would want to put in some stuff about heat sources and temperatures, probably also the effect of differential solar absorption, because thermal flows are what create wind and drafts in the first place. I think it's tricky but doable, and since its a static problem maybe not terribly computationally expensive.
To make it worthwhile, though, I would think you would want to see underground vapors and oxygen deprivation, dusts and such, modeled in a 'better' way than current miasmas are. Otherwise there's not much point in doing the work...IMHO, anyway.