Iris writes non-CF-compliant polar stereographic netCDF files (wrongly named grid mapping attribute)

[TonyP-BAS]

🐛 Bug Report

When loading a polar stereographic netCDF file, Iris correctly interprets the "standard_parallel" attribute of the polar stereographic grid mapping variable, storing this in the "true_scale_lat" attribute of the PolarStereographic coordinate system (line 456 of iris/fileformats/_nc_load_rules/helpers.py).

But when saving such a cube to netCDF, Iris writes this value to an attribute of the grid mapping variable called "true_scale_lat", not "standard_parallel", as required by the CF convention (https://cfconventions.org/Data/cf-conventions/cf-conventions-1.10/cf-conventions.html#polar-stereographic).

This looks to be a defect in line 2150 of iris/fileformats/netcdf/saver.py, which reads:

                    cf_var_grid.true_scale_lat = cs.true_scale_lat

Where I would have expected:

                    cf_var_grid.standard_parallel = cs.true_scale_lat

The consequence is that the netCDF written by Iris is not CF-compliant, and cannot even be read back correctly by Iris due to the lack of the required "standard_parallel" or "scale_factor_at_projection_origin" grid mapping attributes, meaning that the cube ends up lacking horizontal coordinates or a coordinate system.

Happy to provide example data file and code to reproduce the bug, or any other information that might be useful.

Environment

• Iris Version: 3.3 or later (i.e. one with polar stereographic support)

[pp-mo]

Thanks @TonyP-BAS
I don't know how it slipped through exactly, but clearly you are right.
From the recently-added coverage scan, these lines are not exercised by any tests whatever, so no wonder.
( I would link, but it seems the CodeCov site is currently down ☹️ )

Do you feel able to submit a PR for this yourself ?
I think the code change is probably just as simple as you suggest, but we should definitely add a test.
The simplest I can think of is an integration test to check that Iris can save and re-load one of these and retrieve a coord-system with all the same parameters.
Obviously, it's a shame that tests don't exist for other coord-system types ... but I don't think there is any obligation to fix that here+now.

[pp-mo]

I don't know how it slipped through

But it appears that I merged it #4773.
😳