[MITgcm-devel] Fwd: Submeso Parameterization

[Dimitris Menemenlis]

I am forwarding Baylor's e-mail to Devel list so that I can reference  
it in CVS check-in.  D.

> Hi Guys,
> I've spent a couple of days working out the submeso
> parameterization.  Take a look and see what you think. I haven't
> checked it in, but I have made it the default ;-).   I've been testing
> it on the tutorial_global_oce_latlon with an override on the mixed
> layer depth to make sure a few gridpoints reside in the ML.  (Without
> this, the submeso parameterization doesn't do anything in that test).
> It could certainly use more extensive testing in runs that really
> have a KPP mixed layer that is sometimes deeper than the first 3
> gridpoints.  It should b
>
> At this point, I could really use some more eyes on it.  Especially
> JMC's since he finds my off-by-one errors so effectively  ;-).  After
> his approval, it can be checked in, since Galen and others are
> waiting.
>
> I've coded it as an extension to gmredi.  I've included the old
> version of gmredi I was working from to diff against.  It was at the
> head of the tree about a month ago, so it may be current.
>
> The parameterization works in either the bolus advection form or the
> tensor form.  The former is 'native', as that's what I tested offline
> in Fox-Kemper & Ferrari.  However, the other form should work, too,
> but it requires the use of GM_ExtraDiag, since there won't be any
> magic cancellation between the skew and redi fluxes in this case
> (which I've flagged in gmredi_check.F and made a switch for in
> gmredi_readparms.F).
>
> I've also done some basic additions to diagnostics, and a
> data.diagnostics entry is below.  At the moment, I've coded so that
> diagnostics for fluxes in bolus or tensor form are always calculated
> accurately, regardless of which form is used in the forward run.
> (That is, even when GM_Bolus is false, GM_SM_Psix will not be zero,
> even though it won't be used by the model integration).
>
> It is possible to run into CFL violation problems due to the scaleup
> factor.  This can be ameliorated by making GM_SM_Lmax smaller or
> GM_SM_Lf bigger.
>
> It is easy to do control runs by setting GM_SM_Ce to zero (which turns
> off the parameterization).
>
> Happy hunting, and I will collect my beers from those of you in
> Boulder this week...
>   -Baylor
>
>
> &diagnostics_list
> diag_mnc=.TRUE.,
> frequency(1) = 1800.,
> fields(1,1) = 'GM_KuzTz','GM_KvzTz',
>               'GM_PsiX ','GM_PsiY ',
> filename(1) = 'GM',
> frequency(2) = 1800.,
> fields(1,2) = 'SM_KuzTz','SM_KvzTz','SM_KrddT',
>               'SM_PsiX ','SM_PsiY ',
>               'SM_ubT  ','SM_vbT  ','SM_wbT  ',
> filename(2) = 'SM',
> &