Switching to single variable files in the ESM UM netcdf conversion

Feedback needed - vote below on proposed changes

ESM1.5/6 ocean model output has single variable files with names like
ocean-2d-surface_temp-1monthly-mean-ym_0101_01.nc which have a year of data.

The atmosphere netCDF output currently has two files for each month, one with all monthly means and one with all daily variables.

We are considering switching this to single variable files, each with a year of data. This would give faster loading for many diagnostics and would integrate better with the intake catalog.

The standard output settings would create ~ 200 files/year similar to the ~250 for the ocean so this would not be an unreasonable increase in inode usage.

File naming would depend on whether the variable names are also switched (https://forum.access-hive.org.au/t/proposed-changes-to-variable-names-in-um-netcdf-conversion/4599). If the ocean model convention is followed it could be something like atm-ts-1monthly-mean-ym_0101_01.nc. (See Essential diagnostics · Issue #374 · ACCESS-NRI/access-om3-configs · GitHub for the full discussion of the ocean model file naming.)

An alternate suggestion was to keep multi-variable files but to have some sort of logical grouping, e.g. atmosphere, radiation, land, BGC. The grouping could be done in the UM STASH configuration but mapping to sensible group names for the netCDF files might be fragile.

Suggestions welcome.

This needs to be settled before the ESM1.6 CMIP7 piControl starts so please vote before Friday 20th June.

  • No change - keep multivariable files
  • Single variable files
  • Single variable files but only if also switch to meaningful variable names
  • Some alternate multivariable grouping
0 voters