This proposal is for a standard_name for the real/imaginary
component, i.e. a way of coding complex data in a standard way --
talking about amplitude/phase is a total red herring, and fourier
transform was merely an example, again, not the proposal.
The fact remains that all software that is treating the data as
complex will want both components -- if one is only interested in one
part, then we are not talking about complex data any more.
This discussion should go on to spherical harmonics, and EOFs, as
Bryan suggested, not repeating the same Fourier example over and over.
Maybe we could pick up on Fourier after talking about some of the
other examples.
I have non-standard structures for both spherical harmonics and EOFs.
EOFs are fairly straightforward, a dataset with a number of related
variables, e.g.
http://iridl.ldeo.columbia.edu/expert/SOURCES/.CAC/.ssta/%7BY/cosd%7D%5BX/Y%5D%5BT%5Dsvd/
as opendap
http://iridl.ldeo.columbia.edu/expert/SOURCES/.CAC/.ssta/%7BY/cosd%7D%5BX/Y%5D%5BT%5Dsvd/dods
as netcdf
http://iridl.ldeo.columbia.edu/expert/SOURCES/.CAC/.ssta/%7BY/cosd%7D%5BX/Y%5D%5BT%5Dsvd/data.cdf
In this case the time series retain the original units, and the
structures are normalized to r.m.s. 1. So the time series can be
tagged by inheriting the tags of the original data, since all we have
done is exchanged the spacial dependence (X,Y) for a modal dependence
(ev).
This would suggest new standard_names for the eigenvalue independent
variable, and the normalized structures.
As for spherical harmonics, it is an important example, because I am
pretty sure almost everyone who stores them thinks of them as a
compression of the data, or, as I put it earlier, a basis
transformation. On the other hand, my experience is limited to
triangular truncation and the velocity/temperature basis distinction,
and the structure I use essentially tags it as such -- a modeler would
be better suited to discussing what would be representations of the
entire family of spherical harmonics.
Benno
On Fri, Jul 9, 2010 at 2:58 AM, Jonathan Gregory
<j.m.gregory at reading.ac.uk> wrote:
> Dear Bryan
>
> I may have overemphasised the point that you might be interested in only one
> of the amplitude and phase. My main point is you might want to express a FT
> as amplitude and phase, rather than real and imaginary. Since amplitude and
> phase don't have the same units, I think it would be against CF conventions
> to give them the same standard name and put them in the same data variable.
> The same treatment should apply to real and imaginary. But you might not agree
> with that argument on grounds of consistency.
>
> Although it may look clunky, I think using standard names to distinguish the
> two parts of the FT is actually clearer - more self-describing - than using an
> index coordinate variable with arbitrary values (such as 0=real 1=imaginary -
> or would it be the other way round - and what if they are amplitude and phase?)
> to do it. That's another reason for my preference.
>
> Best wishes
>
> Jonathan
> _______________________________________________
> CF-metadata mailing list
> CF-metadata at cgd.ucar.edu
> http://mailman.cgd.ucar.edu/mailman/listinfo/cf-metadata
>
--
Dr. M. Benno Blumenthal? ? ? ? ? benno at iri.columbia.edu
International Research Institute for climate and society
The Earth Institute at Columbia University
Lamont Campus, Palisades NY 10964-8000?? (845) 680-4450
Received on Fri Jul 09 2010 - 07:59:07 BST
