Proposal for diaptr

[marti]

Please find enclose a proposal for diaptr.F90 :

• Correct a bug when ln_mskland=.true. ;
  
• Separate frequency of computation to frequency of writing (add n_ptr_wri to namelist) ;
  
• Add a namelist flag to calculate only transport, or to add zonal means and meridional stream functions (ln_diaznl) ;
  
• Add in ouput files zonal surfaces on which each zonal mean is computed, to allow mean computation for larger depth slices ;
  
• Correctly masks transports and meridional stream function south of 30S in subasins for ORCA ;
  
• Correct latitude definition (only for ORCA2 and ORCA05) ;
  
• Compute zonal means at T points, not V points ;
  
• Add computation for all subasins ;
  
• Add EIV part to transports and meridional stream functions when available (CMIP conformance).

Tested in IPSL coupled model in ORCA2 configuration. All flags tested.

In the future :

• separate computation frequency for transport and zonal mean ?
  
• do we need overturning and gyre separation ?

Olivier

[marti]

diaptr.F90

[marti]

Example for namelist. Only &namptr is changes

[smasson]

commit a first version given by O. Marti, see changetset:1340
WARNING: Not working in MPP...

[rblod]

commit a version from O. Marti working in mpp when jpji*jpnj=jpnij, the number of the writing processor should be reviewed

[rblod]

As it is, flio_dom_set is called even in mono processor case, and there is a mix between nidom and nidom_ptr:
Fix :

• add nidom_ptr
  
• call flio_dom_set in mpp case only, and initialise nidom_ptr by hand otherwise

[charris]

The order of calculations has changed in the new version of the routine such that the v_msf_glo used to calculate the overturning heat/salt transport is now the stream-function rather than the zonally averaged meridional velocity.

[charris]

I still think there is a problem here (see comment above) because in [1340] the order of calculations was changed relating to pht_ove, v_msf_glo etc.

[charris]

Fixed version of diaptr

[charris]

I've attached a new version of diaptr which fixes the problem I described above by moving the position of the stream-function calculations. We can discuss at the next teleconference whether this can go straight onto the trunk.

I'm also attaching a fixed version of traadv_tvd.F90 required for the ptr diagnostics to work properly.

[charris]

Fixed version of tvd advection code for poleward transport diagnostics

[charris]

Code changes agreed with Olivier Marti and committed to trunk in [1877].