Hello CF Having considered the feedback to my previous posts, I think that the approach of defining a new method of 'recording the spacing of the original data' by adding the capability to define a cell method 'domain' based on ancillary variables is a positive additional capability, not currently supported by CF, with utility in a number of scenarios.
If there are no strong objections I will raise a trac ticket proposing this enhancement to CF. thank you mark ________________________________ From: CF-metadata [[email protected]] on behalf of Hedley, Mark [[email protected]] Sent: 04 November 2013 10:56 To: CF metadata Subject: Re: [CF-metadata] Cell methods when there are no coordinates Hello Jonathan et al The current syntax is clear about using the 'name' to reference a dimension for the cell method. In the case I have described I am wary about creating length one coordinates for realization, experiment_id, source and institution, it is not clear to me how to do this clearly. This is why I am interested to explore the idea of extending the cell_method syntax to allow ancillary data to be referenced from a cell_method. I see this as additional semantics, enhancing the current capability. I have re-read 7.3.2 Recording the spacing of the original data and other information specifically To indicate more precisely how the cell method was applied, extra information may be included in parentheses () after the identification of the method. This information includes standardized and non-standardized parts. Currently the only standardized information is to provide the typical interval between the original data values to which the method was applied .... This makes me wonder whether my example could be handled by extending this section. Something like: To explicitly define the domain over which the statistic was calculated, the syntax (domain: var_name var_name) may be used, where each 'var_name' is an ancillary variable of the data variable. This would enable the cell method for the example I posted to be defined: sh_sd:cell_methods = "leadtime: mean (interval 6 h) ensemble: mean (domain: experiment_id source realization institution)" ; Does this look like a reasonable approach to take? thank you mark ________________________________________ From: CF-metadata [[email protected]] on behalf of Jonathan Gregory [[email protected]] Sent: 01 November 2013 15:21 To: [email protected] Subject: Re: [CF-metadata] Cell methods when there are no coordinates Dear Mark In the cell_methods attribute, the entries have the form "name: method", as you know. The name usually identifies a dimension (see first para of 7.3), and it doesn't matter if the dimension doesn't have a coordinate variable. I agree with you that cell_methods is the best way to describe an ensemble statistic, such as ensemble mean, and there have been previous discussions about doing this over previous years. I think your second example should look something like this. dimensions: ensemble = 1 ; variables: int realization(ensemble); realization:long_name = "Number of the simulation in the ensemble" ; char experiment_id(ensemble,STRINGLENGTH); experiment_id:long_name = "Experiment identifier" ; char source(ensemble,STRINGLENGTH); source:long_name = "Method of production of the data" ; char institution(ensemble,STRINGLENGTH); institution:long_name = "Institution responsible for the forecast system" ; float sh_sd(ensemble, time, latitude, longitude) ; sh_sd:units = "1" ; sh_sd:standard_name = "specific_humidity" ; sh_sd:cell_methods = "leadtime: mean (interval 6 h) ensemble: mean"; sh_sd:coordinates = "experiment_id source realization institution" ; CF doesn't give any guidance about the contents of the aux coord vars after aggregation; the data processing software will have to decide how best to represent the ranges of the collapsed ensemble axis. For numerical aux coord vars such as realization above you could provide bounds. Best wishes Jonathan _______________________________________________ CF-metadata mailing list [email protected] http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
_______________________________________________ CF-metadata mailing list [email protected] http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
