dtatem & dtasal : cleaning - move the call from step to tradmp - merge in 1 module

[gm]

In previous version of NEMO, the 3D field of T and S were used for (1) internal damping to levitus (key_tradmp) and (2) for sst and sss restoring. This was the reason why dta_tem and dta_sal where called at the begining of step before the surface fluxes computation and the tra_dmp call.

With the new surface module, the optional restoring to sst ans sss now uses another data file containing only the 2D temp and salinity data required for the sea surface restoring. The read of 3D field is now only required if the key_tradmp is defined (and in the initialisation, when starting from levitus). Therefore, the call to dta_tem and _sal can now be moved in tradmp module. Furthermore, it is suggested to combine the two modules (dtatem & dtasal) into a single module as they are almost identical. Finally, in dtatem & stasal, fldread generalised to 3D field and on-the-fly interpolation can be introduced.

Required action : - move the dta_tem dta_sal call from step.F90 to tradmp.F90

• merge dtatem dtasal sal modules in 1 single module named dta_t_s.F90

• generalised fldread for 3D fields created and used in dta_t_s.F90

• on-the-fly interpolation introduced in dta_t_s.F90 (both for initial state and internal damping)

[molines]

Although this ticket was open a long time ago ... I think that we must also consider that the initial data are not necesseraly the data you will use for tradmp.

You may start from an old run, or you may want to use an annual climatology for damping in the deep ocean (eg Gouretski), not available as monthly climatology and therefore not usable for initial conditions. In recent DRAKKAR simulation, in dtatem dtasal module we just add a dta_tem_0 and dta_sal_0 routine which is called from istate for initial condition.

The enhancement proposed in this ticket may easily integrate the requirements i have just exposed.

[smasson]

I completely agree with you. This part is not clean (I was discussing on this with Gurvan yesterday).
Luckily, Clement Bricaut implemented the 3D version of fldread that was the necessary step to change/merge dtatem and dtasal. work is in progress...
see: https://forge.ipsl.jussieu.fr/nemo/browser/branches/DEV_r1784_3DF

[gm]

merge of dtatem-dtasal + dynanical allocation+suppression of key_tradmp

[gm]

Starting from the v3.3beta, a merge of dtasal-dtatem into dtatsd.F90 has been made together with the use of dynamical allocation and suppression of key_tradmp.

The resulting code is provided in attachment (GM-dtasal_dtatem_merge_TRADMP.zip). It has been compiled in changeset:2436 (one of the last of v3.3beta), but not further tested. One of the 2011 LOCEAN task will be to include this in the reference version. 
Gurvan