[MITgcm-support] atmospheric exp broken ?
Martin Losch
mlosch at awi-bremerhaven.de
Thu Oct 16 02:33:13 EDT 2003
Jean-Michel,
as far as I remember, the default for useNHMTerms=.true. stems from the
historical fact, that before the inclusion of this flag, the terms
where always included. That's a bug, but we didn't want to change the
results of the verification experiments. So I strongly agree with you:
Change the default for useNHMTerms and use them only where appropriate.
None of the hydrostatic experiments should have these terms.
Martin
On Thursday, October 16, 2003, at 02:07 AM, Jean-Michel Campin wrote:
> Hi,
>
> I choose to set "gravitySign" automatically, according to
> the vertical coordinate that is used.
> I was thinking that such a minor modification will not affect
> the output of the model.
> Instead, it appears that all the atmospheric experiments that
> use the flux form are "broken":
>
> Y Y Y Y 6 16 16 16 16 16 16 16 16 16 16 16 16 6 5 7 6 FAIL
> adjustment.128x64x1
> Y Y Y Y 4 7 7 9 8 7 6 8 7 4 4 6 5 4 3 3 5 FAIL
> aim.5l_Equatorial_Channel
> Y Y Y Y 8 8 7 8 8 7 7 6 7 4 5 6 6 5 4 3 6 FAIL
> aim.5l_LatLon
> Y Y Y Y 6 11 8 9 8 16 16 16 16 4 5 5 5 5 5 5 5 FAIL
> hs94.128x64x5
> Y Y Y Y 4 9 7 8 7 16 16 16 16 4 5 4 4 5 5 22 5 FAIL
> hs94.1x64x5
>
> because the default is to use non-hydrostatic metric terms
> (however, not very clear for me what non-hydrostatic metric terms are
> in a one layer adjusment problem with no momentum advection like in
> adjustment.128x64x1) that were "wrong" because the vertical
> coordinate was assumed to point in the wrong direction.
>
> I would recommend:
> 1) to change the default value of useNHMTerms
> and to set it to "false" instead of "true".
> (was only introduced recently and was not in release.1)
> 2) for experiments where it make sense, to turn it on
> so that the results are unchanged.
> 3) for the other experiments where it's meaningless (1 layer) or
> wrong (atmosphere using mom_fluxform), to regenerate new output
> with the (new) default useNHMTerms=F.
>
> comments and suggestions welcome.
>
> Jean-Michel
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://dev.mitgcm.org/mailman/listinfo/mitgcm-support
More information about the MITgcm-support
mailing list