[CF-metadata] Handling data from land surface models with sub grid scale tiling

Jonathan Gregory j.m.gregory at reading.ac.uk
Thu Mar 13 02:15:51 MDT 2008


Dear Martin

You're right, the CF standard doesn't explicitly say how to deal with this, but
there are some precedents and provision. The currently open trac ticket 17
(http://cf-pcmdi.llnl.gov/trac/ticket/17) is concerned with related issues.
That has not been accepted yet, but almost three weeks have passed since any
further comment was added and I don't think there are any outstanding
disagreements, so I hope that it will be accepted. The up-to-date version of
the proposed changes to the convention is near the end of the ticket.

The assumption in that ticket is that you would have a dimension for surface
type (tile) as you have done. If there is a (1D) coordinate variable it could
only be an index, like you have, but you can also include a string-valued
auxiliary coordinate variable i.e. a char variable (tile,maxstringlength) with
standard name of area_type to identify the surface types. The possible values
of area_type will be specified by a list maintained with the standard name
table, and added to by proposing them to this email group, like stdnames.

CF doesn't require the dimensions to be in any particular order, but for
compatibility with COARDS it is generally recommended to put
(time,depth,lat,lon) at the right (in CDL or C order), so the tile
would then come on the left. However that may not be convenient or efficient
for you.

I hope that helps. Best wishes

Jonathan



More information about the CF-metadata mailing list