Loading in multiple forecasts in netcdf format creates multiple cubes rather than a single cube

[daflack]

Describe the bug

Cubes in netcdfs when read in will create a cube list rather than a cube with forecast_reference_time and forecast_period. This led to needing ensure_aggregatable_across_cases but this operator will not work with the difference plots. Therefore, we need to use the "slice and dice method" in read_cubes to produce relevant single cubes.

How to reproduce

Steps to reproduce the behaviour:

1. Load multiple forecasts at the same time in netcdf format
2. See they are loaded in CubeList

Expected behaviour

Cubes from multiple forecasts should come as a single cube.

Environment

• Version: [e.g. from cset --version]
• Browser (if UI issue): [e.g. Chrome, Firefox]

[daflack]

Due to the usage of read_cubes "slice and dice" is not the solution for this problem.

[jfrost-mo]

I've just been in the AVD surgery to determine the cause of the non-merging coordinates in NetCDF files. It comes from where the NetCDF file has the time coordinate as an extra dimension in the underlying cube, where PP files have each field being a scalar one, so essentially it is already broken apart.

I've got a suggestion of where to fix it in CSET, and considerations around how to make it reasonably efficient, though it will have some performance cost still.

[jfrost-mo]

This may also be useful: https://scitools-iris.readthedocs.io/en/stable/further_topics/controlling_merge.html