[MITgcm-devel] switching seaice_obcs from legacy to devel

[Martin Losch]

Hi Dimitris,

sitting vis a vis with Jean-Michel at his kitchen table, here's my (really: our) take on this:
1. we don't want to use the legacy code anymore and in fact most (all) of us don't, so it's better to test the code that is being used.
2. having very small ice thicknesses (10e-3) with 100% does not appear very physical, so searching for this would mean searching for a problem in code that we actually want to get rid off.

So let's go ahead and use the non-legacy code in this experiment without further worrying about the differences.

Martin

On Dec 15, 2012, at 12:41 AM, "Menemenlis, Dimitris (3248)" <Dimitris.Menemenlis at jpl.nasa.gov> wrote:

> Dear MITgcm sea ice developers, as requested I am working on switching
> verification/seaice_obcs from legacy to devel code.
> 
> With #undef SEAICE_GROWTH_LEGACY
> I get some rather large differences in the output of the 3 experiments:
> 
> Y Y Y Y> 2< 2  7  5  5  5  9  7  4  1  1  2  2 16  2  3  2  0  2  0  0  0  4  3  3 FAIL  seaice_obcs
> Y Y Y Y> 2< 2  7  5  4  5  9  7  4  1  1  3  2 16  1  3  2  0  2  0  0  0  3  3  3 FAIL  seaice_obcs.seaiceSponge
> Y Y Y Y> 3< 2  7  5  5  5 10  7  4  2  1  3  3  1  3  3  3  0  2  0  0  0  4  3  3 FAIL  seaice_obcs.tides
> 
> I have traced this difference back to large differences in "AREA" between legacy and devel branches.
> In attached figure I plot AREA, HEFF, and Qsw after 6 model time steps for the legacy
> and devel branches an for the difference.
> Note that in region of very thin ice (lower right quadrant of panels),
> the legacy code gives ~100% ice cover, while the devel code give ~0% ice cover.
> This has big consequence for shortwave radiation.
> 
> I have following questions:
> 
> - Do you know what is causing this difference?
> 
> - Are there compile-time or run-time flag options that can reduce this difference?
> 
> - Shall we just check in the new output.txt as is?
>   For the things that are being tested in seaice_obcs, this difference is probably not important,
>   and the devel treatment is probably more realistic in terms of shortwave forcing.
> 
> Dimitris Menemenlis
> 
> <legacy.jpeg>
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel