Making initial condition files for icosa_lmdz.exe from
an LMDZ start_archive.nc file

What you need to do/know before:
---------------------------
Run a dynamico simulation (with startphy_file=false for example) with the desired resolution to get
start.nc and startfi.nc. These files will be used as reference (rename as *_ref.nc) in the script.
You can use the files named *_no_start to have an example of files to run without startfiles

Job_mpi is the file you need to run start_archive2icosa on Occigen

Warning: the input start_archive.nc file must include SSO parameters in this version, which is only available with a modified start2archive: start2archive_SSO

This script is currently under development and still needs to be finalized.
So you must be careful to adapt the script to your own needs and replace hardcoded variables.
Any suggestion for improvement is welcome.

In a nutshell:
--------------
1) input file start_archive.nc needs some minor adaptations:
- variable "nivsigs" needs be renamed "sigs" (XIOS requires CF-compliant files) (RV: No neeeded anymore)
ncrename -vnivsigs,sigs start_archive.nc
- input file start_archive.nc need be converted to NetCDF4 format
nccopy -k 4 start_archive.nc start_archive_nc4.nc
2) script "compile" can be used to compile main tool start_archive2icosa.f90
   as well as rearrange_startphy.f90
3) to run, start_archive2icosa needs input files:
- "start_archive_nc4.nc" : from step 1
- "start_icosa_ref.nc" : a start.nc file from an icosa_lmdz.exe run (contains
  all the necessary information about the icosaedral grid at that resolution)
- "startphy_icosa_ref.nc" : a startfi.nc file from an icosa_lmdz.exe run (contains
  all the necessary information about colums from physics)
- "surface_nc4.nc" : Mars topography file (also needs to be in NetCDF4 format)
- "iodef.xml" : information about field names and grids for XIOS
NB: interpolation of HR topography requires lots of memory; see e.g. job_mpi
job script where 2 nodes (56 cores) are requested but only 28 processes
are run.
4) start_archive2icosa outputs files "start_icosa_prefinalize.nc" and "startfi_prefinalize.nc"
  require a couple of extra modifications. Script "icosa_finalize.bash" does
  the required changes and generates files "start_icosa.nc" and "startfi.nc" that
  can be used as inputs by icosa_lmdz.exe
5) When using these start one must add in run_icosa.def the extra flag
etat0_start_file_colocated = true

Additional comments:
--------------------
- script "compile" assumes XIOS is to be found in "/scratch/cnt0027/lmd1167/adelavois/models/code/XIOS"
  to be adapted to your own settings
- "nccopy" is a standard netcdf tool available via e.g.:
module load intel/17.0 intelmpi/2017.0.098 hdf5/1.8.17 netcdf/4.4.0_fortran-4.4.2
- "nco" tools (ncrename, ncks, ...) are available via e.g.:
module load intel/17.0 openmpi/intel/2.0.1 hdf5/1.8.17 netcdf/4.4.0_fortran-4.4.2 nco/4.4.6
- in start_archive2icosa.f90, rearrange_startphy.f90, iodef.xml and
  icosa_finalize.bash input/output file names are hard-coded.
  Modify/adapt with caution!