[CF-metadata] Question about the 'time' coordinate

From: Thomas Lavergne <thomasl>
Date: Tue, 28 Oct 2008 15:11:46 +0000

Hi all,

In the 'units' attribute of any time variables, there must be a mention of the epoch:
:units = "seconds since 1978-01-01". Does this epoch needs to be a string?

For example, if I have a time variable which indeed is a "delta time". Its unit is "seconds since
reference_time". And this "reference_time" is changing from one file to the next. For example, the
reference_time of a daily product might be the_same_day_at_00utc which might be a value provided by
a 'time' variable, a global attribute or even a dataset's attribute.

For example:

netcdf nctime {
dimensions:
         time = 1 ;
         station = 3 ;
variables:
         int time(time) ;
                 time:long_name = "reference time of product" ;
                 time:units = "seconds since 1978-01-01 00:00:00" ;
         float dt(time, station) ;
                 dt:long_name = "delta time for measure" ;
                 dt:units = "seconds since reference_time" ;
                 dt:_FillValue = -1.e+10f ;
                 dt:valid_min = 0.f ;
                 dt:valid_max = 86400.f ;
                 dt:coordinates = "station" ;

// global attributes:
                 :Conventions = "CF-1.3" ;
                 :date_and_time = "2008-10-01 0:0:0"
}

Is there a valid way, within CF, to make reference_time point to:
1) the value of the 'time' variable?
2) the value of a global attribute (e.g. 'date_and_time')?
Can the concept of "ancillary variables" be adapted / extended to this?

Of course, I could myself compute the "date string" associated to the value of the "time" dimension
and use its value in the units of "dt". But it is not optimal as, in my understanding, attributes to
variables are designed for software (not human) and such a software would have to decode a
stringdate twice;

Note that this issue arises also when we provide a value as "mean_value + delta", whatever the unit
is. Do we have a way to say "hPa around daily average" and "daily average" being defined somewhere else?

And sorry if this is a FAQ or has already been discussed... I browsed the archive but did not find
it (applied to time).

Thanks for help,
Thomas
Received on Tue Oct 28 2008 - 09:11:46 GMT