Changing unit dimensionality in depth integration pre-processor about esmvalcore

[valeriupredoi]

can now use the convert_units preprocessor step 🍺

[bouweandela]

Related to #505

[schlunma]

I don't think we can close this, convert_units (hopefully) doesn't convert units from mol m-3 to mol m -2.

[valeriupredoi]

why not? It's a valid CF unit, the cube doesn't care if it's cows / vultures as long as they are CF units, it's CMOR that will be barking at you, but you're on your own there

[schlunma]

>>> from cf_units import Unit
>>> a = Unit('kg m-3')
>>> b = Unit('kg m-2')
>>> a.convert(1, b)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/miniconda3/envs/mlr/lib/python3.7/site-packages/cf_units/__init__.py", line 1907, in convert
    (self, other))
ValueError: Unable to convert from 'Unit('kg m-3')' to 'Unit('kg m-2')'.
>>> 

[valeriupredoi]

ValueError: Unable to convert from 'Unit('mol m-3')' to 'Unit('mol m-2')'. 🤦‍♂️

[ledm]

We shouldn't be using iris convert here, that isn't what happenning here. Iris convert is for stuff like Celsius to Kelvin, not Celsius to heat. The convert function should not be able to change the dimensions of the units.

The issue here is that the depth_integration preprocessor changes the units, dimensions and the dataset standard name. It's a very simple calculation, but non-trivial to resolve it, as the output may not necessarily exist in the CMOR tables.

For instance, you may want to know the water column total oxygen, in mmol per meter squared. I bet that field doesn't exists a standard variable.

[bouweandela]

@ledm That should no longer be a problem, because we moved the cmor check to right after the data is loaded in #743, so there is no CMOR check anymore after the depth_integration function is applied.