Operations should recognize auxiliary variables

[forman]

Selected ECT operations should be able to correctly recognize auxiliary variables associated with a measurement variable. For example

• uncertainty variables provided for a measurement may be considered when deriving measurement statistics
• uncertainty variables may be resampled using a different methodoloy
• when extracting a measurement variable from a dataset, its auxiliary variables may be automatically extracted too

[JanisGailis]

http://cfconventions.org/Data/cf-conventions/cf-conventions-1.1/build/ch03s04.html
http://cfconventions.org/Data/cf-conventions/cf-conventions-1.1/build/apc.html

[JanisGailis]

Also:
http://cfconventions.org/cf-conventions/v1.6.0/cf-conventions.html#cell-methods

There is also something called NetCDF-U meant to be a CF compliant standard for geo-spatial uncertainty information:
https://www.researchgate.net/publication/258780653_NetCDF-U_-_Uncertainty_conventions_for_netCDF_datasets

However, all the internet resources for NetCDF-U seemed dead, so I don't believe it's an active effort.

[JanisGailis]

Soil Moisture doesn't have a standard name attribute, hence its uncertainy doesn't follow the scheme laid out in CF conventions.

[JanisGailis]

See also #114

[JanisGailis]

Operations that need to work with ancillary variables (propagate uncertainties), should depend on using the 'ancillary_variables' attribute of an xarray dataset, as per CF Conventions.

To mitigate the situation that many datasets don't abide by this out of the box, there could be an operation add_ancillary that would create the associations in the dataset. Then, this can be run before using operations that depend on ancillary variables. Note that the ancillary variables themselves are supposed to have the original dataset's standard_name in their standard_name fields. Also, uncertainty variables have well defined standard_names that should be used, or added.