# Condensation Rate: Devil in a Detail

When gas disappears somewhere in the atmosphere, local pressure is lowered and a compensating air inflow from the surrounding areas is initiated. In our paper “Where do winds come from?” (M10) we derive the magnitude of a stationary horizontal pressure gradient $\partial p/\partial x$ that is associated with water vapor condensation — the process by which the vapor gas molecules are packed into a thousand of times smaller liquid volume and thus effectively disappear from the atmosphere.

There has been much critical discussion in the blogosphere and further on the ACPD web site of our Equation 34 for condensation rate $S$ that is key to the presented derivation. At various times and places, including Section 4.2 of our paper, it was pointed out that if one formulates $S$ in terms of water vapor mixing ratio $\gamma_d \equiv N_v/N_d = p_v/p_d$ one obtains $\partial p/\partial x = 0$. If one instead uses the relative partial pressure of water vapor, $\gamma \equiv N_v/N = p_v/p$, a horizontal pressure gradient is obtained that appears to be so significant as to substantiate the claim for a dominant role in the whole planetary dynamics. (Here $N_v$, $N_d$, $N = N_v + N_d$, $p_v$, $p_d$, $p = p_v + p_d$ are molar density and pressure of water vapor, dry air and air as a whole, respectively.)

A typical value of water vapor partial pressure $p_v$ in the lower atmosphere is around 1-3 per cent. This means that the mixing ratio and relative partial pressure $\gamma_d$ and $\gamma$ differ very insignificantly. Reckoning up the preceding criticisms, our second referee Dr. Isaac Held referred to this difference as to “a detail”. It is the purpose of this note to show that the $\gamma/\gamma_d$ dichotomy is precisely the devilish detail that, as it increasingly appears, is responsible for the fact that the condensation-induced dynamics has not received the attention it deserves.

Consider the stationary continuity (mass conservation) equations written for dry air and water vapor (Eqs. 32 and 33 in M10): $\nabla N_d \mathbf{v} = 0$,      $\nabla N_v \mathbf{v} = S$.      These equations only tell us that the dry air mass is conserved, while the vapor mass may be conserved or it may be not: there can be a local source or sink of vapor $S$. Irrespective of the existence/nature/magnitude of the vapor sink/source $S$, the above equations can be combined with use of elementary algebra such that their left-hand parts take various forms. In his review Dr. Held chose $N_d \mathbf{v} \nabla (N_v/N_d) = S$, which is equivalent to

$\mathbf{v} (\nabla N_v - \gamma_d \nabla N_d) = S$.    [1]

In the two-dimensional circulation considered in M10 the velocity vector is the sum of the horizontal and vertical components, $\mathbf{v} = \mathbf{u} + \mathbf{w}$. We also consider a horizontally uniform surface temperature, which dictates a constant saturated pressure of water vapor, such that $\mathbf{u} \nabla N_v = 0$.     [2] (If water vapor is not saturated, this assumption corresponds to a horizontally uniform surface temperature and constant relative humidity.) Combining [1] and [2] we obtain

$\displaystyle \mathbf{u} \nabla N_d = \left(S - S_d\right)\frac{1}{\gamma_d}$,     where $S_d \equiv \mathbf{w} \left(\nabla N_v - \gamma_d \nabla N_d \right)$.    [3]

Equation [3] has two important implications for any given horizontal velocity $u \ne 0$. First, it shows that when $S = S_d$, the horizontal density gradient $\partial N/\partial x = \partial N/\partial x = 0$. Second, it shows that if $S$ and $S_d$ differ by a small relative magnitude of the order of $\gamma_d \ll 1$, this magnitude is multiplied by a large relative magnitude $1/\gamma_d \gg 1$ to determine the horizontal density gradient $\partial N/\partial x$.

We should emphasize that the two above conclusions are unrelated to any ideas about what the condensation rate could look like. The value of $S$ in [3] is unknown. Equation [3] shows that any minor difference of the order of $\gamma_d$ in the theoretical formulation of $S$ — whatever the latter might be — is not a detail but is the zeroth order term in determining the horizontal density and pressure gradients associated with vapor condensation.