Query about the expected behaviour of interface_fluxes in non-neutral conditions

[mmr0]

Hi,

I have been reading the awesome docs on interface fluxes, but am perplexed by the bottom panel of this figure figure:

As far as I understand the example, the atmosphere specific humidity is set to be equal to the surface saturation humidity, so there should be no vapor fluxes. So I would expect the sign of the air-sea temperature difference ΔT to set the sign of the buoyancy flux. Also as per my understanding and also the text:

The relationship between the relative air-sea state and turbulent fluxes is modified by the presence of buoyancy fluxes -- "destabilizing" fluxes, which stimulate convection, tend to increase turbulent exchange, while stabilizing fluxes suppress turbulence and turbulent exchange.

So I'm confused as to why a large positive ΔT of 20K corresponding to a stabilising buoyancy flux would lead to an increased drag coefficient?

Thanks in advance for any comments/clarifications!

[glwagner]

This does look wrong to me. The stability function plot also makes it clear that the fluxes should decrease if the stability parameter increases.

Something that is missing is a plot of the stability parameter that comes out of the flux computation. That might be one place to start for figuring out what's going wrong here...

cc @simone-silvestri

[simone-silvestri]

Thanks for reporting this @mmr0 and sorry that it took me some time to address this.
I think you uncovered a very subtle bug in the flux computation:
we use gustiness to enhance the velocity difference, this gustiness is diagnosed from the buoyancy flux, where the gustiness velocity is modeled with

Uᴳ = β * cbrt(Jᵇ * h_bℓ)

In the above, β is a gustiness free-parameter (O(1)), while h_bℓ is the height of the atmospheric boundary layer and Jᵇ is the bouyancy flux. This velocity is used to enhance the velocity difference between ocean and atmosphere

U = sqrt(Δu^2 + Δv^2 + Uᴳ^2))

so even when there is no large-scale velocity difference between the ocean and the atmosphere, there is always a small scale (turbulent) velocity that generates turbulent exchange of heat and vapor.
The problem is in the sign of Jᵇ, when this is negative, the boundary layer should stabilize, so the gustiness velocity should be zero theoretically, however, since we square it, the gustiness increases, predicting higher gustiness for a stable boundary layer.

I have opened an issue from this discussion which is definitely a bug.

[mmr0]

Awesome thanks @simone-silvestri for looking into this - the new plot in #748 looks much better!