scale_factor and add_offset for dimension variables

• To: Netcdf group <netcdfgroup@xxxxxxxxxxxxxxxx>
• Subject: scale_factor and add_offset for dimension variables
• From: Philippe Poilbarbe <Philippe.Poilbarbe@xxxxxx>
• Date: Fri, 27 Apr 2007 10:19:26 +0200

Hello,

We are defining  a new product file for satellite data which
will be launched (with an altimeter instrument) and we have
data referenced by their date (date of measurement).
The date has a precision of 1 microsecond and its unit
is in seconds since 2000-01-01 00:00:00.
So for a launch in 2010 and a life cycle of 5 to 15 years
we need numbers with at least 16 significant digits.

We have defined (for example) a 'date' dimension with
its dimension variable.
This variable is (by default) of type double and in
order to reduce the number of significant digits needed
we did try to specify an add_offset attribute which will
contain the date of the first data in the file (at a precision
of 1 second => 7 digits). So the date of the first
measurement in the file is stored as a value comprised
between 0 and 0.999999 (6 digits).

But when we try to visualize the data with general
tools (like ncBrowse), this attribute is not taken
into account when attached to a dimension
variable.

After rereading the CF convention it seems that there
is something about this but it is not clear (for me). It
also has a reference to the netCDF User Guide which
does not contain any special thing for scale_factor
or add_offset attached to a dimension variable.

So I would be glad if somebody can tell me if using
these attributes attached to a dimension variable is
considered as a bad usage (and possibly why) or
if this is a problem/bug of the general tools.

Thanks,
Philippe.

--
Philippe Poilbarbe
CLS, Space Oceanography Division

==============================================================================
To unsubscribe netcdfgroup, visit:
http://www.unidata.ucar.edu/mailing-list-delete-form.html
==============================================================================