[CF-metadata] example for CF docs

From: John Caron <caron>
Date: Fri, 15 Oct 2004 09:36:37 -0600

Steve Hankin wrote:

>John Caron wrote:
>
>
>
>>Hi Steve:
>>
>>CF has actually started down the path of specifying the projection
>>parameters (see Appendix F), which we are taking from FGDC. So I was
>>really asking for a double check on a "best practice" of how all the
>>pieces should fit together, and an example added to section 5.6. We are
>>currently writing code that creates CF netcdf datasets out of GRIB, so
>>it seemed like a good time to double check.
>>
>>
>
>Hi John,
>
>Well, I guess my face should be a bit red that I was not even aware of Appendix
>F! Obviously, I am not the right person to comment on whether you have applied
>the Appendix F syntax correctly. My discomfiture is offset by the good news
>that Appendix follows exactly the strategy of drawing from an existing standard
>(FGDC). Great minds do think in the same gutter, I guess.
>
>
"Consistency is the hobgoblin of small minds, or mid-sized minds seeking
common funding". Thoreau

>
>
>>BTW, as a philosophical matter, I would consider the specification of
>>the projection as fundamental, and the specification of the explicit
>>coordinate in lat, lon as redundant, because you can figure out the
>>values from the projection, but not the other way around. This is not to
>>say that I disagree with the decision to require the lat,lon values but
>>not require the projection, however. But i would (and do) recommend
>>adding the projection info. If the software knows how to use it, it
>>makes visualization of the grid easier, since model grids are regular in
>>the projection coordinates, and so they constitutes a kind of "natural"
>>coordinate system. So I would advocate that CF recommends adding
>>projection info.
>>
>>
>
>I agree with your conclusion (that CF recommends adding projection info) but
>qualify it by saying "where projection information is applicable". Only a
>fraction of the many tri-polar, coastline following, etc. curvilinear grids in
>use by modelers represent named projections. The dependence on named,
>standardized projections has been a serious limitation of (say) GRIB from the
>modelers' perspective. From this pov the projection name is a convenience
>label that should be used to indicate special, standardized instances of
>curvilinear coordinate systems.
>
> - steve
>
>
I agree! How about:

"CF recommends adding projection info when applicable, in the
standardized ways described in Appendix F and Section 5.6".
Received on Fri Oct 15 2004 - 09:36:37 BST