Clear sky upwelling LW not equal to total sky upwelling LW at surface?

[khartig]

I am confused why the clearsky upwelling longwave flux at the surface (FLNSC - FLDSC) is not equal to the total sky upwelling longwave flux the the surface (FLNS - FLDS). There are regions where the total sky upwelling flux is over 100 W/m2 and/or 1.5+ times more than the clear sky.

I looked through rrtmg_lw_rad.f90 and rrtmg_lw_rtrnmc.f90 and the only obvious difference I found at the surface is in the specular reflection of downwelling longwave, but since the surface emissivity is ~90% the reflected radiance should only be on the order of 10% at most outside of desserts and things, which would be 10's of W/m2 rather than the 100+ I'm seeing. This is all with cesm2.1.3 with CAM6, F2000climo compset at f09_f09_mg17 resolution. Any help in interpreting RRTMG and these variables would be greatly appreciated!

 

[brianpm]

Hi Kara,

I think you are using an incorrect sign convention.


In rad_rrtmg_lw, you can see the definition of the net surface fluxes:

flns(:ncol)  = uflx (:ncol,1) - dflx (:ncol,1)
flnsc(:ncol) = uflxc(:ncol,1) - dflxc(:ncol,1)

So to derive the upwelling surface flux:

FLUS = flns + flns
FLUSC = flnsc + flnsc

While we do not have these as output fields, they are easy to add to confirm.

To do that, you can add the following to the subroutine radiation_init in radiation.F90:

call addfld('FLUS'//diag(icall), horiz_only, 'A', 'W/m2', 'Upwelling longwave flux at surface', sampling_seq='radlwsw')
call addfld('FLUSC'//diag(icall), horiz_only, 'A', 'W/m2', 'Clearsky upwelling longwave flux at surface', sampling_seq='radlwsw')

and then to get output, you can add the following to subroutine rad_rrtmg_lw in radlw.F90:

call outfld('FLUS', uflx (:ncol,1), pcols, lchnk)
call outfld('FLUSC', uflxc(:ncol,1), pcols, lchnk)

(around line 217 or so).

I just did this and wrote instantaneous, time-step output for one day of an F2000climo run.

When I look at that output using python (with xarray and numpy):

>>>> (np.absolute(ds['FLUSC']-ds['FLUS'])).max()

array(0., dtype=float32)

So the actual upwelling fluxes are identical at all grid points at all the timesteps for that one day. The derived upwelling flux shows only a negligible discrepancy:

>>>> np.absolute((ds['FLNS'] + ds['FLDS']) - ds['FLUS']).max()
array(6.1035156e-05, dtype=float32)
>>>> np.absolute((ds['FLNSC'] + ds['FLDSC']) - ds['FLUSC']).max()
array(6.1035156e-05, dtype=float32)
>>>> np.absolute((ds['FLNSC'] + ds['FLDSC']) - (ds['FLNS'] + ds['FLDS'])).max()
array(6.1035156e-05, dtype=float32)

 

[khartig]

Thanks for the correction! I saw somewhere that the flux sign convention was positive down, which lead me to believe that net would be up plus down, so I appreciate the clarification.

I'm assuming you meant to write
FLUS = FLNS + FLDS
FLUSC = FLNSC + FLDSC
(you had net twice on the right-hand side)

 

[khartig]

It looks like SW fluxes have the opposite sign convention from LW; rrtmg_sw sets the net flux with down minus up,
swnflxc(i) = swdflxc(iplon,i) - swuflxc(iplon,i) swnflx(i) = swdflx(iplon,i) - swuflx(iplon,i)

 

[brianpm]

Correct. The conventions, I think, are set so the net fluxes are (generally) positive.

 

[brianpm]

Thanks for the correction! I saw somewhere that the flux sign convention was positive down, which lead me to believe that net would be up plus down, so I appreciate the clarification.

I'm assuming you meant to write
FLUS = FLNS + FLDS
FLUSC = FLNSC + FLDSC
(you had net twice on the right-hand side)

Right!