[MITgcm-devel] Add sea ice surface forcing to pkg/seaice

[Menemenlis, Dimitris (3248)]

Jean-Michel, thank you for looking at this.

I don't want to waste too much of your, Gael's, or Martin's time with this
so I will check in a version of seaice_growth.F that does the same thing
but does not change the verification results.  The code is slightly
longer (and uglier?) than Revision 1.166, but I think OK until more
folks start using snowPrecip as input for pkg/seaice.

Answers to your comments below:

1) lab_sea.hb87 seems very sensitive to truncation errors.
Even before latest changes to seaice_growth, it fails on my laptop.

I am interested in those two conditions:
>  a) if the seaice-fraction is small (e.g., 10^-3), and a snow storm
>  pass by, will all the snow pile up over this little ice fraction ?
>  b) snow over fully open ocean grid-cell ?
so I made sure that they give reasonable-loking results.
Snow over fully open ocean grid does what we expect it to do, i.e.,
cool the ocean surface through negative latent heat release and then,
when SST reaches the freezing point, it starts growing "fresh" ice, provided
SEAICE_VARIABLE_SALINITY and SEAICEuseFlooding are defined.
The sea ice fraction (AREA) is initially small, but it very quickly grows
to reasonable-looking values.  Since we don't know how the ice
concentration should grow in open ocean during a snow storm,
I think that default behavior is OK for now.  The most important
feature is the negative latent heat release, which is modeled
with sufficient accuracy irrespective of sea ice fraction.

I tried to do this:
> c) less urgent: heat conservation and heat budget diagnostics ?
> Would be good to clarify point 2.a before updating the results
> (in case seaice_growth.F needs to be changed again).
correctly, but a second pair of eyes on this would not hurt.
With version 1.167 of seaice_growth.F, all the stuff pertaining
to snowPrecip will be contained within
"IF ( snowPrecipFile .NE. ' ' ) THEN, ..., ENDIF"
so if there is issues with heat conservation,
they will be restricted to the specific case of snowPrecip.

Finally, I notice that:
1) seaice_growth.F does a lot of diagnostic computations, whether they are needed or not.
Shouldn't these computations be contained within
IF ( DIAGNOSTICS_IS_ON('????????',myThid) ) THEN, ..., ENDIF
statements?
2) seaice_growth.F contains lots of constants, e.g., "0.0 _d 0", etc., which could be
replaced with parameters "ZERO", etc., from SEAICE_PARAMS.h
Happy to take care of either of these two changes, if requested to do so.

Dimitris Menemenlis

On Apr 23, 2012, at 6:12 AM, Jean-Michel Campin wrote:

> Hi Dimitris and others,
> 
> few things:
> 1) my impression (after a quick look at your changes) is that 
> with empty snowprecipFile, it's only truncation error changes.
> But still 2 fwd exp are failing (global_ocean.cs32x15.seaice & lab_sea.hb87
> since the criteria on baudelaire with gfortran is 13 digits) 
> which would require to update the output.
> Regarding AD tests, seems OK, don't see any new "fail".
> 
> 2) I don't have time now to check the case where snowprecip is specified
> (+ little time I have is spent checking which testreport output changes)
> and I don't know much about seaice_growth.F ; Can someone (Gael, Dimitris, 
> Martin, ...) check what does snowprecip do in the current seaice_growth.F ?
> In particular:
>  a) if the seaice-fraction is small (e.g., 10^-3), and a snow storm
>  pass by, will all the snow pile up over this little ice fraction ?
>  b) snow over fully open ocean grid-cell ?
>  c) less urgent: heat conservation and heat budget diagnostics ?
> Would be good to clarify point 2.a before updating the results
> (in case seaice_growth.F needs to be changed again).
> 
> Cheers,
> Jean-Michel