Currently we generate separate data folders for dycore, initialization, and physics regression data. This was done because we must generate these datasets in separate runs, and it is not feasible to merge Serialbox datasets. Now that we convert Serialbox data to netcdf, we could merge this data into a single folder.
Merging these folders would avoid needing to specify TARGET values when running tests, which currently can cause tests to pass without running.
This also requires removing our last dependency on Serialbox so that we can stop archiving Serialbox data. The "serialbox" initialization of the driver must be updated to use netCDF data.
Currently we generate separate data folders for dycore, initialization, and physics regression data. This was done because we must generate these datasets in separate runs, and it is not feasible to merge Serialbox datasets. Now that we convert Serialbox data to netcdf, we could merge this data into a single folder.
Merging these folders would avoid needing to specify TARGET values when running tests, which currently can cause tests to pass without running.
This also requires removing our last dependency on Serialbox so that we can stop archiving Serialbox data. The "serialbox" initialization of the driver must be updated to use netCDF data.