Is there a molecular diffusion component to wind? I'm thinking of vapor pressure from terrestrial water: A highly evaporating area in the ocean creates a volume of high relative humidity, the vapor tends to move to less humid air and by molecular friction drags along some air...
Obviously such a force would be pretty much hidden by thermal related components, but as it would sometimes oppose and sometimes work with thermal vectors, I'm wondering if there might be some anomalies explainable thereby.
Is there something like this in existing weather studies/models, or am I misunderstanding something?
 A: This is a rather astute observation and one that is worth discussing! When the composition of a gas changes, as in your example where water vapor concentration is larger in one location than another, that composition gradient wants to equalize. This shows up as a diffusion of concentration and does, in fact, introduce a velocity between the two regions. In the turbulent combustion world we deal with the same things, although our thermal and concentration gradients are much larger than what you see in the atmosphere. When the diffusion of a scalar (like species or temperature) is aligned with the others, it is called gradient diffusion. When they are aligned opposite each other, it is called counter-gradient diffusion. 
Anyway, we can put some estimates to all of this. From my favorite book on turbulence, A First Course in Turbulence, they work through this example (not a direct quote, this is coming from my notes from many years ago):

Consider a room of length $L$ filled with still air all at the same temperature. There is a radiator on one side acting as a heat source [edit: we can also call it a humidor putting out water vapor, the same ideas will work]. Based on the dimensions of the heat diffusion equation, we have:
  $$ \frac{\nabla T}{t_h} = \nu \frac{\nabla T}{L^2} $$
  Therefore, the time scale of the diffusion is:
  $$ t_h = \frac{L^2}{\nu} $$

Okay, so that gives us a time scale of the heat diffusion alone. We can also come up with a velocity scale, which is the distance divided by the time, and so the diffusion velocity would be something like:
$$V_{diff} = \frac{\nu}{L}$$
Now let's look at what that would be for your example. Based on orders of magnitude, $\nu \approx 10^{-5}$ and the length scale of a species gradient in the atmosphere is probably pretty big... let's just assume it is at least on the order of 1 meter, such that $L \approx 1$. 
From that, the diffusion velocity is really low. Like microns per second. I can't imagine that ever showing up in the atmosphere as an important contributor. It does show up in combustion -- there, the length scales are on the order of microns and so the diffusion velocity can be larger than the convective/flame speeds. But for atmospheric flows, it's pretty hard to imagine that wind due to composition differences alone will show up given all of the other dynamics at play.
