Change `BARRA_DIR`/`ERA_DIR` to `source_dir` and take it as a command-line argument

[atteggiani]

Currently the BARRA_DIR or ERA_DIR paths are hardcoded (and set from the ROSE_DATA environment variable).

It would be better for this path to be declared as a command line argument to the hres_eccb.py/hres_ic.py scripts (for example using the --source or --source_dir option.

This would:

• Make the scripts portable. If I have some valid data (same structure) locally, the scripts can still be used by setting the source path via the --source option. Currently the scripts can only be used on Gadi.
• Give more freedom in case the paths to the source files want/need to be changed. We would only need to change the suite (change the --source path) and not the python scripts.

[penguian]

Currently, the scripts can be used for standalone testing, but only if

1. The environment variable ROSE_DATA is defined and is set to a valid directory name,
2. The directory $ROSE_DATA contains the subdirectory etc, and
3. The subdirectory etc contains an appropriately name link pointing to the correct data directory:
   • barra_r2 for replace_landsurface_with_BARRA2R_IC.py
   • era5_land for replace_landsurface_with_ERA5land_IC.py

Admittedly this is clumsy and should be changed, but it doesn't completely block testing.

[atteggiani]

Curently in the RNS:

1. The reanalysis data (era5-land and barra_r2) are symlinked to a directory within the suite run directory ($ROSE_DATA/etc/...). This step is carried out in the install_cold task.
   Their original data reside in:
   era5-land --> /g/data/zz93/era5-land/reanalysis
   barra_r2 --> /g/data/ob53/BARRA2/output/reanalysis/AUS-11/BOM/ERA5/historical/hres/BARRA-R2/v1
2. Then, In the respective tasks, the entry points hres_eccb and hres_ic use that data as a source to replace the data for the restart file

I don't see any other use of the reanalysis data within the suite, therefore I think that its only purpose is to serve as a source for the replacement of the restart file data.

If that's the case, I propose the following:

1. Within the replace_landsurface scripts, replace this portion and this potion with the the original data directories, depending on the chosen --type:

• /g/data/zz93/era5-land/reanalysis for era5land
• /g/data/ob53/BARRA2/output/reanalysis/AUS-11/BOM/ERA5/historical/hres/BARRA-R2/v1 for barra

2. Remove the symlink creation within the suite install_cold task
3. Update the suite based on these changes (no more ROSE_DATA env variable needed for the hres_... tasks, etc.)

I think this would make it clearer where the original data comes from and it can be easily tested (both unit tests and integration tests)

What do you think?
@penguian @engelca