Add reading with XIOS for new files for nitrogen input

[jgipsl]

After inclusion of ORCHIDEE-CN in the trunk, new files are now read. These need to be read with XIOS as done for the other files. This must be done to keep possible running with DYNAMICO.

Related to ticket #469

[jgipsl]

Reading of soilbulk and ph from file soil_bulk_and_ph.nc is done in changeset [6159].

Note that the reading (for this file) can only be done if the sechiba restart file is not read (RESTART_IN=NONE).
If the sechiba restart exist but is incomplete, XIOS will not be able to read the variables. If this case appeares, the model will exit wiht ipslerr_p.
This is because that the reading of the file must be activated or deactivated before the restart file is read, before the model knows that a variable might be missing.

[jgipsl]

Difference of soilbulk and ph while reading with IOIPSL compaired to XIOS

[jgipsl]

Files read from slowproc_Ninput has now also been adapted to be read by XIOS, see commit in the trunk rev [6519]. Reading with XIOS will be done if XIOS_INTERPOLATION=y. If no reading with XIOS will be done (if there is no need for reading or if IOIPSL is used), the corresponding files and fields are deactivated from context_input_orchidee.xml. This is done from a new subroutine called slowproc_xios_initialize_ninput. The reading of the fields are still done in slowproc_Ninput.

[jgipsl]

There were one bug in the above commit which has been corrected here: [6521]. The mask was not read correctly as the name of the variable was not set from the code.

[jgipsl]

Correction needed for current default setup with LMDZOR configuration. [6526]

There are still problems in debug compilation mode.

[jgipsl]

The problem in debug mode was only relanted to reading of soil_bulk_and_ph.nc file. The solution, found by Yann Meurdesoif is commited here [6657].

[jgipsl]

There is still a problem in compilation mode debug at obelix only. It is not clear yet if it is due to reading of the new nitrogen maps or another old file.

[jgipsl]

Still to be done in this ticket:
Only problem left is now when using compilation mode debug at obelix and XIOS_INTERPOLATION=y. It is the files in read in subroutine slowproc_Ninput which make error at obelix. Test using revision 6700 from the trunk has been done but I do not know why this problem only at obelix (irene and jean-zay are ok).

Only error message I can see is the following:

Remapping...
In file "attribute_template_impl.hpp", function "void xios::CAttributeTemplate<T>::checkEmpty() const [with T = int]",  line 78 -> On checking attribute with id=ni_glo : data is not initialized

Note that in the ORCHIDEE trusting, the test O8d activates reading with XIOS. This test now always fails at obelix but is ok at irene and jean-zay. At obelix, this test has been deactivated from the final status of the trusting (to avoid the trusting being failed all the time).

This issue must be solved before changing to have XIOS_INTERPOLATION=y by default.

[jgipsl]

Another problem (corrections done but not tested):
In slowproc_Ninput, the variables SUMveg_max and SUMmanure_pftweight are calculated only if XIOS_INTERPOLATION=n but they are used further below for all cases of 'Nmanure' (both xios_interpolation=y and n).

Note that the default set up do not read file Nmanure. Instead it reads Nmanure_cropland and Nmanure_pasture. Therefor the standard experiment set up never goes into this bug.

The calculation of these intermediate variables have been corrected in the trunk but this case is still not tested [7598].

[jgipsl]

Decided on todays ORCH meeting: remove case with Nmanure not used anymore, from the trunk and from ORCHIDEE_3 => Josefine

[jgipsl]

Nmanure was removed from the trunk [7604]. It will not be done in ORCHIDEE_3 as it is only a cleaning.

Note that the following 2 lines in orchidee.def in ORCHIDEE_OL experiment set up is now useless but without any harm:

Nmanure_FILE = NONE
Nmanure_VAR = Nmanure

It will not be cleaned right now for the fear of problems with incompatibilty if using an older version of ORCHIDEE with the new version of ORCHIDEE_OL.

[jgipsl]

The problem mentioned above as Still to be done, is treated in a seperate ticket #807 . This ticket can be closed now if trusting is green.

[jgipsl]

Trusting is green at irene.