[CF-metadata] Encoding Errors on variables in CF

[Jonathan Gregory]

Dear Ag and Bryan

> I agree with Bryan here, what we mean by "error" will vary significantly
> between different datasets.

In that case I don't think it is appropriate to have standard_names which use
the same word ("error") or phrase to label these quantities. It would be better
to say more precisely what the quantity is.  When data is produced it must
be possible to say how the error estimate was derived in each case, and there
may be a choice of methods, so we should choose standard names that indicate
what has been done.

I think "uncertainty" would be a better word than "error". I know error often
means uncertainty, but it can also mean bias or mistake!

If we had some concrete examples it would help us to decide what to do.

Best wishes

Jonathan