Nesting of ancillary_variables attribute?

[frankeye]

Is this supported, or a good idea, i.e.

var a:ancillary_variables = "var_b"

var b:ancillary_variables = "var_c var_d

Moreover, if the var_b and var_c, var_d are 'quality_flag's ?

[davidhassell]

Hello @frankeye,

This is not currently supported by the cf data model. In the data model diagram, the FieldAncillary construct is synonomous with a CF-netCDF ancillary variable, and we can see that only a Field construct (i.e. CF-metCDF data variable) can contain it.

That's not the end of the story, though ... Any variable in a CF-netCDF file can always be viewed as a data variable in addition to any metadata role it may have, simply by choosing to ignore any other variables that may reference it.

So in the wholly legal following:

    float var_a ;
        var_a:ancillary_variables = "var_b" ;
    float var_b ;
        var_b:ancillary_variables = "var_c var_d" ;
        var_b:foo = "bar" ;

when we read var_a as a data variable, var_b is designated an ancillary variable and so the var_b:ancillary_variables attribute is "ignored" - rather, it is retained, but only with the same non-standardised status enjoyed by var_b:foo.

However, when we read var_b as a data variable, var_b:ancillary_variables is now standardised and var_c and var_d are read as ancillary variables.

But what we can't currently do is let var_b relate to var_a whilst simultaneously var_c and var_d relate to var_b.

I would be very interested in the use case you have for this, to understand if there is a case for extending CF in this direction.

Thanks,
David