[MITgcm-devel] diagstats_mnc_out.F

[Jean-Michel Campin]

Hello Ed,

I have a few remarks about the mnc version of the diagStat output file:

0) The numerical values seems OK, and it's good !
  I did not check all the fields, but there is no obvious reason
  to have a problem in the .nc file and not in the ASCII output file.

1) _lv_ & _vi_ : I changed my mind. It would be better to have
*in case of 3D field: _lv_ for per level statistics ;
 and without _vi_ (= just the variable name + suffix: _min,_max,_vol,_ave,_std)
 for global (all-levels) statistics ;
*in case of a 2D field, it should not be _lv_ but also just the variable
 name + suffix.
<- I checked in those changes.

2) a rather stupid thing:
 I would prefer to have the global statistics written before
 the per-levels one (like in the ascci file), so that it comes
 first when I do an ncdump.
<- I checked in those changes.

3) When I do ncdump, I would prefer to see, e.g. ETAN_lv_vol
 with 2 dimensions (Time, Region) instead of 3 (Time, region, Zd000001).
<- I don't know how to do this.

4) vertical coordinate would be useful.
<- I saw the "DIAGST_MNC_NEEDSWORK"

5) I think it would be nice to put the right time (computed from
 myTime, frequency & phase) to correspond to the midle of the
 time period (if averaged) or the time of the snap-shot.
<- not urgent

6) would better not put any "units" for _vol field, since
   "udiag" is not the right units for this variable
  (But it's hard to know exactly what they are).
  And the minimum, maximum, average & standard-deviation
  are missing in the description.

7) I need only 1 output file (and I get only 1 ascii file) but
 get plenty of .nc files (1 per tile), all identical except bi,bj
 in the header.
<- This is anoying and needs to be fixed.
   And since there is no "tile" information, we can drop the
   tile number in the file-name suffix.

Have a nice WE.

Jean-Michel