closea behaviour if closea_mask not present

[mathiot]

Context

I found the behaviour of closea a bit dangerous (ie potential source of bug for the user).
If closea_mask is not present in domain_cfg, ln_closea has no effect and only a small comments is written in ocean.output:

    ==>>>   closea_mask field not found in domain_cfg file.
            No closed seas defined.

It looks very easy to miss this information and run a long simulation thinking the closed sea are removed (for example).

Suggestion

I think it is safer to set a integer switch:

• nn_closea = 0 ! => nothing is done (closea_mask defined or not)
• nn_closea = 1 ! => closed sea are removed using closea_mask (error raised if variable are missing)
• nn_closea = 2 ! => special treatment for closed sea using fields defined in domain_cfg.nc (error raised if variables are missing)

Concerning the removal of closed sea:
As improvement, we could used a 2d flood filling algorithm to do the job (ie detect the main ocean and mask everything else). It will remove the need of a closea_mask and the definition of all the hard coded indexes in DOMAINcfg. Similar work as already been done in 3d in UKMO/dev_isf_remapping_UKESM_GO6package_r9314@9513 for the removal of subglacial lake beneath the ice shelf. The only extra input needed in the namelist will be the coordinates in lat/lon of a seed in the main ocean (default could be (0,0) to work with all eORCA and ORCA configuration).

[mathiot]

Move in 2019 WP as part of ENHANCE-03_PierreMathiot_DomcfgTools.

[mathiot]

Fixed by the merge 2019. Will be included once the trunk updated to the merge branch