Output combining script

[ccarouge]

The TRENDY branch is not using MPI for parallelisation but splitting the simulation in a number of regions and running a separate simulation for each region. The script handles the splitting at the start and the merging of the simulations outputs at the end. However, the script merging the outputs depends on a bespoke package that is not commonly used.

This issue is to try and redesign the script to avoid this dependence.

[Whyborn]

This is being worked on in this branch in the TRENDY repository.

[Whyborn]

The merge script on the mentioned TRENDY branch now works. There are some "cosmetic" differences between the outputs generated by the old and new scripts:

• The old output includes "latitude/longitude" variables (gridded versions of the "y/x" variables) and "local_lat/local_lon" variables, which are masked versions of "latitude/longitude". The new output only includes the vector "y/x" outputs.
• The old output didn't actually mask the inactive points on the grid, just filled with a pseudo-fill value, not the NetCDF _FillValue. The new output actually has these points masked.
• The biogeophysics output (bios_out_cable_<stage>nc) doesn't include the soil_carbon_pools and plant_carbon_pools dimensions, which are not included on any biogeophysics variables.

[ccarouge]

@Whyborn , wouldn't we want the latitude and longitude variables in the new outputs as well? Seems like that would be useful (I'm not certain what the x and y variables contain).

[Whyborn]

Well depends what the goal is- consistency with the previous outcome, or results that are more in line with the standards. The x and y variables are what should be called lon and lat- vectors describing the grid. In the original script, latitude and longitude are 2D variables of size (nlat, nlon), describing the latitude and longitude of every point. Note that what the script does is consistent with the outputs of the CABLE_POP-TRENDY branch- the x and y variables are the variables that should be lon and lat in a standard-conforming dataset.

[ccarouge]

Ah I see. I agree that adding the lon and lat variables seems a waste of storage then.