thank you for the feedback Steve, you raise some interesting points
With regard to the history or cell_method, I think I would prefer to use the cell_method, as it has a well defined syntax construction which fits well with defining the operation I have performed.
To make this more explicit, I have updated my example to maintain the ensemble dimension I have aggregated over, with length 1. This defines a data dimension, without a coordinate defining it. I think this better represents the notion of a cell within my data over which a cell_method has been applied.
https://cf-pcmdi.llnl.gov/trac/wiki/aggregateExampleMH
As stated, I would like to define the conceptual range over which the aggregation took place, in my cell_method entry. To do this, I think it makes sense to reference the relevant ancillary variables.
I hope this clarifies the case slightly.
many thanks
mark
________________________________
From: Steve Hankin [steven.c.hankin at noaa.gov]
Sent: 25 October 2013 17:15
To: Hedley, Mark
Cc: CF metadata
Subject: Re: [CF-metadata] Cell methods when there are no coordinates
Hi Mark,
Volumes of documentation have been written about cell_methods that I have not kept up with, so this response comes with implicit caveats. Maybe this response is "coming from left field". But here goes ...
In your ensemble mean example (the second/result example on your Wiki) the ensemble axis is no longer an axis of the variable sh_sd. The ensemble mean operation is a uniform property over the entire domain of the variable sh_sd, rather than a mean that was calculated over a unique coordinate range represented by each cell of an axis. Rather than using cell_methods, shouldn't the ensemble mean be documented through a "history" variable, or some other machinery that describes the lineage of calculations that went into creating this variable?
- Steve
P.S. In the result dataset, in which you have included all of the ancillary variables that describe the ensemble (a nice touch), the string "ensemble:mean" becomes a (potentially) nice, complete description of the operation that has been performed -- clean and self-describing.
================================================
On 10/25/2013 2:55 AM, Hedley, Mark wrote:
Hello CF
I am interested in defining cell methods where there are no coordinates.
I have a set of cases which I do not think fit the patterns in the current conventions.
I suggest that we provide further capability for defining cell method instances.
I have prepared a use case on the Trac wiki, to illustrate one of the cases that is driving this:
https://cf-pcmdi.llnl.gov/trac/wiki/aggregateExampleMH
Am I right that there is not a currently available capability for defining the result data set in this case?
I think this case and many others may be addressed by adding the following text to the conventions document, appending to:
7.3.4. Cell methods when there are no coordinates
the text:
Where the cell method refers to a nominal cell, which is not described by a data dimension, ancillary variables may be referenced by a cell method to indicate the concepts the aggregation took place over.
To define this, the syntax:
"ancillary_variable_name_1: ancillary_variable_name_2: cell_method_name"
is used.
In this case, referenced ancillary variables may be extensive, with multiple values, but they must be invariant with respect to the data variable: they may share no data dimensions.
_______________________________________________
CF-metadata mailing list
CF-metadata at cgd.ucar.edu<mailto:CF-metadata at cgd.ucar.edu>
http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <
http://mailman.cgd.ucar.edu/pipermail/cf-metadata/attachments/20131028/79c12220/attachment-0001.html>
Received on Mon Oct 28 2013 - 03:41:20 GMT
