Investigate the "must-be-filled rule" for PARAMETER_UNITS

[delphinedobler]

Item raised by JPR:
PARAMETER_UNITS refers to the unit of PARAMETER_ACCURACY and PARAMETER_RESOLUTION as explicitely stated in the UM.
Currently, there is a must-be-filled rule in the FileChecker for PARAMETER_UNITS, however, when PARAMETER_ACCURACY and PARAMETER_RESOLUTION are not filled in, this has no meaning, and thus there is no sense in constraining the metadata creator to fill it.

So the rule could be changed as follows:

• if PARAMETER_ACCURACY and/or PARAMETER_RESOLUTION is filled in, then PARAMETER_UNITS must be filled. -> alert level = rejection
• if PARAMETER_ACCURACY and PARAMETER_RESOLUTION are not filled in, then PARAMETER_UNITS must not be filled. -> alert level = warning (for the moment)

any comment back are welcome

[jprannou29]

First remark: never reject the file whatever the result of the check is.

Second remark:
In some specifications we don't have only a value in the PARAMETER_ACCURACY/PARAMETER_RESOLUTION.
Example
param = 'DOXY';
paramSensor = 'OPTODE_DOXY';
paramUnits = 'umol/kg';
paramAccuracy = '8 umol/kg or 10%';
paramResolution = '1 umol/kg';

In such case PARAMETER_UNITS may be empty or not.

My recommendations would be:

• if PARAMETER_ACCURACY and/or PARAMETER_RESOLUTION is filled in with only one numerical value, then PARAMETER_UNITS must be filled. -> alert level = warning
  

• if PARAMETER_ACCURACY and PARAMETER_RESOLUTION are not filled in, then PARAMETER_UNITS must not be filled. -> alert level = none or warning ? (to say that the information is useless)
  

[JohnEGilson]

I'm a bit confused, about the general discussion and interpretation of the variables. I see the wording in the UM that Delphine refers to. However, my interpretation would not be that this wording forms a link between only the 3 PARAM variables (units, accuracy and resolution). My interpretation has always been that all 5 variables are linked. If I extend the thought that only the 3 PARAMS are linked, then would it be permissible to have the PARAMETER (e.g. PSAL) in the various netCDF have different units then is indicated in the meta netCDF (e.g. PARAMETER_UNITS)?

My main point is that I don't agree that having UNITS filled and ACCURACY/RESOLUTION not filled has no meaning. UNITS will provide the information of the units throughout the netCDF family. Yes there is an attribute under the variables in the profile netCDF, but can that differ then the meta netCDF PARAMETER_UNITS? Is my understanding incorrect?

As far as the more detailed discussion about mandatory filling...I believe the current rule supports my thinking above.

[delphinedobler]

Thanks for your comment John,

I will raise this point at the next AVTT meeting to clarify the meaning then, and discuss whether the UM shall be updated or the FileChecker rule. This may also have echos with Annie's comment on the definition of 3.2 metadata format_version.