[CF-metadata] Encoding Errors on variables in CF

From: Brian Eaton <eaton>
Date: Thu, 20 Mar 2003 18:47:15 -0700

All,

The two methods being discussed for linking a variable with some associated
measure of quality are standard names and a linking attribute. I am in
favor of a linking attribute. I'm happy with the "error_variable" proposed
by Bryan. I think "data_quality" might also be suitable and possibly more
generic. I anticipate that the variable pointed to by a linking attribute
will have other yet to be determined attributes and/or a standard name
attached which will more precisely describe the error/uncertainty/quality.
But until agreement is reached on these issues the "comment" attribute is a
good way to encode this information. (Ag, thanks for pointing out the typo
in section 1.3. It's fixed in draft5.)

It appears from the discussion that what is meant by
error/uncertainty/quality is not well defined and this makes incorporation
into standard names difficult. Also, since these are very general concepts
they can apply to almost every name in the current table. This is similar
to the case of statistical quantities. We chose not to incorporate the
statistical method into the standard name for a couple of reasons:
1) to avoid the resulting explosion in the size of the name table, and
2) to provide more information about the nature of the statistic (e.g., its
   domain) by using the "methods" attribute.

Brian

-----------------------------------------------------------------
Brian Eaton | email: eaton at ucar.edu
Climate Modeling Section |
National Center for Atmospheric Research |
P.O. Box 3000, Boulder, CO 80307 |
Received on Thu Mar 20 2003 - 18:47:15 GMT