Hi Jonathan,
Our thinking is the same on this. Thanks.
By the way it is a real use case and I think it is scientifically
defensible: I have 11 full years of data and I first compute a
climatological annual cycle, yielding 12 climatological months. Then I
compute the climatology for each of the seasons from the monthly
climatology. This allows me to use all the data in the 11 years. I
especially would want to do it this way if I only had 2 or 3 years of
data because I wouldn't want to omit any of it.
[Of course you wouldn't be able to do this if you were calculating say a
trend of individual seasons; in that case you clearly wouldn't want to
consider partial seasons.]
best wishes,
Karl
On 2/11/16 12:20 PM, Jonathan Gregory wrote:
> Dear Karl
>
> I agree with Seth that this isn't anticipated by the design of the convention,
> which assumes that the mean over seasons is composed of a number of complete
> seasons. Is this a real use-case, with a DJF mean made by including JF from one
> NH winter and D from another? It seems a bit odd to me. I wouldn't compute a
> monthly mean of anything if 2/3 of the days were missing in the month. But,
> following this logic, your original choice of "1999-12-1" to "2011-3-1" is OK,
> and it "just happens" that Dec 1999 and Jan-Feb 2011 are actually not included,
> as if they were missing data. Their omission is not recorded by the climatology
> bounds but, equally, if Dec 2005 or Jan-Feb 2008 (for example) were missing
> when computing the mean you would not know about it from the climatology
> bounds. So perhaps it doesn't matter.
>
> To spell out exactly which months were used, it would be necessary to record
> also the time coordinate and bounds before the collapse. In various tickets we
> have discussed but not agreed a convention for doing that, as extra info.
> Alternatively you could record it as unstandardised info as a comment in () in
> the cell_methods, as you note.
>
> Best wishes
>
> Jonathan
>
> ----- Forwarded message from Karl Taylor <taylor13 at llnl.gov> -----
>
>> Date: Wed, 10 Feb 2016 15:37:01 -0800
>> From: Karl Taylor <taylor13 at llnl.gov>
>> To: cf-metadata at cgd.ucar.edu
>> Subject: [CF-metadata] How to build CF-compliant seasonal climatology when
>> data begins within a season
>> User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.9; rv:38.0)
>> Gecko/20100101 Thunderbird/38.5.1
>>
>> Dear CF community,
>>
>> In representing the seasonal climatology based on data available for
>> the period January 1, 2000 through December 31 2010, what would be
>> the correct climatology_bounds?
>>
>> climatology_bounds = "1999-12-1", "2011-3-1",
>> "2000-3-1", "2010-6-1",
>> "2000-6-1", "2010-9-1",
>> "2000-9-1", "2010-12-1" ????
>>
>> I would note that this seems to capture the idea that we are
>> reporting seasonal means, but it also seems to indicate that this is
>> based in part on data from Dec. 1999 and Jan.-Feb. 2011, when it
>> isn't. Is this the best I can do? [Of course the convention can
>> never tell us if data are complete in forming a climatology. If 1
>> year were missing, this would not affect the attributes.]
>>
>> I suppose the in the cell_methods attribute ("time: mean over days
>> time: mean over years" I could add non-standardized information (as
>> permitted by CF), for example: "time: mean over days time: mean
>> over years (with data from the period 2000-1-1 to 2011-1-1)"
>>
>> thanks for any suggestions,
>> Karl
>>
>>
>>
>> _______________________________________________
>> CF-metadata mailing list
>> CF-metadata at cgd.ucar.edu
>> http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
> ----- End forwarded message -----
> _______________________________________________
> CF-metadata mailing list
> CF-metadata at cgd.ucar.edu
> http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
Received on Thu Feb 11 2016 - 13:49:53 GMT
