Double counting kv_slow in MOM_vert_friction.F90

[Hallberg-NOAA]

I believe that the contribution of visc%Kv_slow is being double counted at open boundaries by the code starting at about line 1196 of MOM_vert_friction.F90, and that visc%Kv_shear is being double counted in its contribution to a_cpl when visc%Kv_slow is associated. Simply setting Kv_add in this block of code instead of incrementing it should correct the problem, but doing so could change answers.

I do not believe that this code is in use by anyone at GFDL, but someone at NCAR, like @gustavo-marques, should perhaps look into what this might be doing in solutions that do use visc%Kv_slow.

[gustavo-marques]

@Hallberg-NOAA, thanks for reporting this. You are correct, visc%Kv_slow is double-counted at open boundaries.

We set ADD_KV_SLOW=False since a fatal error that avoids the double-counting is issued when this option is used in conjunction with use_KPP = True. ADD_KV_SLOW was meant to be a temporary workaround to allow diffusivities due to slowly varying processes to be passed to the solver. Later in the development process, visc%Kv_slow was added into visc/%Kv_shear so that KPP could match the interior/surface diffusivities.

I believe we should remove this entire block of code (and corresponding code related to ADD_KV_SLOW). If we decide to go this way, I can make these changes on dev/ncar, make sure answers do not change and then obsolete ADD_KV_SLOW. We would still want visc%Kv_slow for diagnostics purposes.

[gustavo-marques]

Closed via #972