Jonathan Gregory wrote:
>Dear John
>
>Like Karl, I wonder why these lat,lon variables are functions of time. Of
>course that increases the storage requirements, although I don't think it
>affects the basic argument.
>
yes i wondered about that, too. what i do know is that is how the
variables are being stored in the current (non CF) netcdf file. i will
ask the WRF group if its really time dependent.
>
>Even if there are 16 ancillary coordinate variables, I do think that is the
>right thing to do, because it means the file will be capable of being
>interpreted by any application which recognises the coordinates attribute.
>If you don't use this mechanism, it is unlikely that most applications will
>be able to locate the data.
>
i agree with keeping output files simple so that standard clients can
read them. i think thats an important goal. however, it can conflict
with other goals, like keeping files small, and capturing the data in
the way that the model actually uses it.
>
>So the question is whether you also need to record the relation of these
>various grids with extra (redundant) metadata. Maybe it would be helpful to
>an application, but I'm not sure it's really necessary, is it? As you said
>earlier, the methods used are quite straightforward: you interpolate one
>field onto the grid of another one, or maybe you interpolate two fields
>onto a common grid which uses x from one and y from the other. To do this,
>no special machinery is required. All the program needs to know is which
>grids are to be used as target for the interpolation. I would argue that's
>not metadata, but a convention for how the data should be used - it's not
>a property of the dataset but of the program.
>
what seems to me to be missing is "which grids are to be used as target
for the interpolation". i dont understand how that information is
already there if you have seperate, non-connected coordinate systems.
the fact that the coordinates are related i would argue is a very
important property intrinsic to the data, not the display.
>
>Best wishes
>
>Jonathan
>_______________________________________________
>CF-metadata mailing list
>CF-metadata at cgd.ucar.edu
>http://www.cgd.ucar.edu/mailman/listinfo/cf-metadata
>
>
Received on Wed Jan 28 2004 - 09:59:57 GMT
