PP save fails with bounded `model_level_number` coordinate

[DPeterK]

If you save to PP with a model_level_number coordinate present that includes bounds, the PP saver code fails. This can be easily reproduced by collapsing the model_level_number coordinate in a cube:

cube = iris.load_cube(iris.sample_data_path('uk_hires.pp'), 'air_potential_temperature')
no_vertical = cube.collapsed('model_level_number', iris.analysis.MEAN)
print(no_vertical.coord('model_level_number').bounds)
[[ 1 19]]

iris.save(no_vertical, 'test.pp')
ValueError                                Traceback (most recent call last)
...
/.../iris/lib/iris/fileformats/pp_save_rules.py in _vertical_rules(cube, pp)
    656     # Single depth level (non cross-section).
    657     if (mln_coord is not None and
--> 658             not mln_coord.bounds and
    659             depth_coord is not None and
    660             not depth_coord.bounds):
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

The problem code is this logical test in the PP save rules. Looks like the bounds test is not set up to correctly handle arrays.

[pp-mo]

I just looked into this.
Obviously it is a bug, simply because if(coord.bounds) is not a safe thing to do, given the slightly peculiar behaviour of arrays in boolean tests.
This code, and the bug, was is inherited unchanged from the old text-base save rules, in #2795.

The intent of this code is really to check
if mln_coord is not None and mln_coord.bounds is None, so that is the change we should make IMHO.
On the other hand, if we fix that bug, it still won't make it possible to save this data with its vertical coordinate information : If the model-level-number coord is bounded, then this rule won't trigger, and it won't record any other level info (sigma or level-pressure) either.

The point is, in UM file formats (PP/FF) there is no defined way to encode data that represents a statistic over a set of levels -- unlike over time periods.
We could perhaps stretch a point and save the upper+lower vertical bounds of the sigma and level-pressure, but leave mln = 0. But even that is not actually allowed for in the F3 document : there is no standard way of encoding what this data actually is.

So the bottom line is, if you want to save this you need to 'pretend' that it is something else.
In this case, it will mean setting mln_coord.bounds = None anyway, I think.
That would work (well, it will save something) without needing the bugfix.

[pp-mo]

Given that the bugfix won't address the original reported problem, I don't think this is a priority for the bugfix release, and I'm going to skip it for now (remove from 2.2.1 release milestone).

[github-actions]

In order to maintain a backlog of relevant issues, we automatically label them as stale after 500 days of inactivity.

If this issue is still important to you, then please comment on this issue and the stale label will be removed.

Otherwise this issue will be automatically closed in 28 days time.

[github-actions]

This stale issue has been automatically closed due to a lack of community activity.

If you still care about this issue, then please either:

• Re-open this issue, if you have sufficient permissions, or
• Add a comment pinging @SciTools/iris-devs who will re-open on your behalf.