Dear Jonathan,
Thanks a lot for your clarification. I was not aware about the CF checker. Such validator is a great
idea!
Best regards
Alex
On 09/11/13 21:47, Jonathan Gregory wrote:
> Dear Alex
>
>> I have a question on the time variable of a monthly climatologies collecting e.g. all data from
>> 1900-01-01 to 2000-12-31.
>> As I understand the standard
>> (http://cf-pcmdi.llnl.gov/documents/cf-conventions/1.6/cf-conventions.html#climatological-statistics),
>> I should use the following:
>>
>> dimensions:
>> time=12;
>> nv=2;
>> variables:
>> float temperature(time,lat,lon);
>> temperature:cell_methods="*time: mean within years time: mean over years*";
>> double time(time);
>> time:climatology="climatology_bounds";
>> time:units="days since 1900-1-1";
>> double climatology_bounds(time,nv);
>> data: // time coordinates translated to date/time format
>> time="1900-1-16", "1900-2-16", "1900-3-16", "1900-4-16", ... "1900-12-16";
>> climatology_bounds="1900-1-1", "2000-2-1",
>> "1900-2-1", "2000-3-1",
>> ....
>> "1900-11-1", "2001-12-1";
>>
>>
>> I use the standard Gregorian calendar. Can you confirm that the used cell_methods and
>> climatology_bounds are correct?
> They look correct to me. Although the standard calendar is the default, it
> would do no harm, and might be informative, to include a calendar attribute.
> You can check that your metadata is correct by using the CF checker on the
> CF website.
>
>> I am not so sure about the cell method since the WOA 2009 climatology uses "time:mean within
>> *months* time:mean over *months*" for monthly climatologies
>> (http://data.nodc.noaa.gov/thredds/dodsC/woa/WOA09/NetCDFdata/temperature_monthly_1deg.nc.info).
> That is incorrect. I think the CF checker would report that error. It would be
> good if NODC could correct the error - I wonder if we have any NODC subscribers
> on this list.
>
>> I am wondering if it would not have been easier to define climatology_bounds as an array with the
>> starting and end dates of *all *subintervals (not just the start for the first and the end of the
>> last subinterval). It would be simpler to find out to which time slice of the climatological
>> variable an individual observation would relate to. It would also allow to make climatologies over
>> other cycles (e.g. tides) or non-periodic processes (e.g. ENSO).
> To calculate climatologies over arbitrary periods, you would need the original
> monthly data, wouldn't you, not the monthly climatology. A non-climatological
> dataset has time bounds for each individual time interval, as you describe.
>
> Best wishes
>
> Jonathan
--
Alexander Barth
GeoHydrodynamics and Environment Research, MARE, AGO
University of Li?ge, Sart-Tilman B5,
4000 Li?ge, Belgium
Phone: +32-4-3663664
Fax: +32-4-3669729
Email: a.barth at ulg.ac.be
Received on Tue Nov 12 2013 - 02:20:35 GMT
