[MITgcm-devel] timeave

[Jean-Michel Campin]

Hi Martin,

I haven't checked what's in u/v/t/sFluxtave files (I assume Dimitri
did it right) but regarding TimeAve_full, I don't see the problem.

1) TimeAve_full is set to 0 when TIMEAVE_STATVARS is called from
initialise_varia.F
2) then incremented (+deltaTclock) in TIMEAVE_STATVARS, each time
do_statevars_tave.F is called (i.e. each iteration) from forward_step.F
3) then the normalization is done from timeave_stats_write,
the fields are written to file and then TimeAve_full is reset to 0
when TIMEAVE_STATVARS is called from the end of timeave_stats_write.

I did a title test setting taveFreq=1 and printing TimeAve_full
from timeave_stats_write just before writing u/v/t/sFluxtave and
it seems OK (TimeAve_full = deltaTclock).

And regarding the choice timeave versus diagnostics, it could be a good 
idea to use pkg/diagnostics instead of pkg/timeave, since we have now 
most of the pkg/timeave outputs available with pkg/diagnostics,
plus much more, more flexibility (choice of variables and frequencies),
less memory storage, and better accuracy (state variables
with staggerTimestep for example). And if you use MNC, you should get 
the name+unit and the extra rows (for u,v location) right.
But pkg/timeave will not disappear very soon.

Jean-Michel

On Thu, Mar 03, 2005 at 10:34:21AM +0100, Martin Losch wrote:
> Hi,
> 
> something is fishy with the timeave-pkg. I can't quite figure out why 
> things go wrong, but here is what happens:
> u/v/t/sFluxtave are not normalized by the averaging period, the reason 
> being that timeave_full is zero in timeave_stats_write, whereas 
> timeave_half is correct (as far as I can see).
> timeave_full is computed in timeave_statvars, but if I set taveFreq=1. 
> it is never set (because timeave_stats_write is always called with 
> myiter=niter0, JMC's "dirty" trick, if I remember correctly). however, 
> at the end of timeave_statvars, DDTT = 0.5 deltaTclock, timeave_half = 
> deltaTclock, timeave_full = 0.5 deltaTclock.
> I have no clue what's going on.
> 
> does anybody have an idea?
> 
> Martin
> 
> PS. is timeave still supported or should I learn to used diagnostics?
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://dev.mitgcm.org/mailman/listinfo/mitgcm-devel