[CF-metadata] Separating geolocation/timing co-ordinates from measurement data

From: Stephen Emsley <SEmsley>
Date: Wed, 28 Apr 2010 16:45:09 +0100

Dear Jonathan

They are DEM corrected geodetic latitude, longitude, and altitude spatial co-ordinates as 2D arrays (n_pixels within swatch x n_lines in product) and measurement time as a 1D array (n_lines). So despite the naming the spatial co-ordinates are not Unidata ... and I assume will have to be renamed.

The reason I asked is that I noted the discussion (swatch observational data >> multiple file datasets) last year but on re-reading the was uncertain whether this was work in progress or to be included when the next version of the convention is released.

>From your, V.Balaji's and John Caron's response I guess that this is still under discussion.

Many thanks for your help. I will follow up on the suggestions.

Kind Regards
Steve

-- 
Dr Stephen Emsley??????? ??????????????????????????????????????????????T: +44 (0)1752 764 289
? ARGANS Limited????????????????????????????????????????????????????? ?M: +44 (0)7912 515 418
-----Original Message-----
From: Jonathan Gregory [mailto:jonathan at met.reading.ac.uk] On Behalf Of Jonathan Gregory
Sent: 28 April 2010 15:44
To: Stephen Emsley
Cc: cf-metadata at cgd.ucar.edu
Subject: Re: [CF-metadata] Separating geolocation/timing co-ordinates from measurement data
Dear Steve
>    We have multiple satellite geophysical data products which share the
>    same set of geo-location and timing co-ordinates. To avoid product
>    bloat (e.g. from approx. 30GB to 90GB per orbit) we are considering
>    the possibility of having a single file storing the co-ordinates but
>    we think that this conflicts with our desire to be CF Convention
>    conformant. Is that correct?
It is technically non-conformant but it has already been done in other
applications, including CMIP5 in some respects. I think we need a convention
for this, but we have not agreed one. My suggestion is that it should be OK
to give the name of the coordinate variables in the file where they reside.
Software will have to regard several files as one dataset, in which case it
would be able to find the coordinate variables by name. Are they 1D (Unidata)
coordinate variables, or auxiliary coordinate variable (named by the CF
coordinates attr) that you want to omit?
Best wishes
Jonathan

Received on Wed Apr 28 2010 - 09:45:09 BST