[MITgcm-support] diagnostics: diagstats

[Jean-Michel Campin]

Hi Martin,

regarding point 1:
I think the idea was to be able to retrieve the total Sum
(the one that the model accumulates into qSdiag, before
division by this "vol"). So, it means that it contains
a factor equal to the number of time-step that have been 
accumulated. I don't think it's often used, but it's a
good way to check how the statistics have been computed.
I would recommand to change the documentation (or to add some,
somewhere) rather than to change the output.

regarding point 2:
I think it's not obvious that we want to include, in the statistics,
values that have no influence at all on the model simulation,
that might be valid or crazy, depending on what you provide,
and not obvious to me neither that it's consistent with what is done 
for other stats where we exclude land (i.e. below ocean bottom) 
values (e.g. pressure), mainly because they are irrelevant.
There are also some technical aspects: right now, all stats 
uses some kind of mask; so, if I have an empty tile and
I decide to remove it (put it in blank-tile list), the stats 
don't change. This would no longer be the case without any mask.

Cheers,
Jean-Michel

On Fri, Apr 16, 2010 at 05:05:21PM +0200, Martin Losch wrote:
> Hi,
> 
> two questions about the average/min/max/std/vol
> 
> 1. what I find in DIAGNAME_vol is (at least for the examples that I currently care about: EXFatemp, SIarea, ETAN, THETA, so both 2D and 3D fields) consistently larger by a factor of 10 (exactly) than what I compute from the grid parameters. Is this on purpose? What am I missing? Is this a case of RTFM?
> 
> 2. I was naively assuming that the averaging for atmospheric fields, such as the forcing field EXFatemp would be global, but it appears that EXFatemp is averaged over the ocean only (as an "M" field it is masked with maskC). Is that correct? How can I change that (to get truly global averages)? I saw that for fizhi and land there are special routines, but can I tweak the code to use them?
> 
> Martin
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support