Re: Data on land only

[hankin@xxxxxxxxxxxxxxxxxxx (Steve Hankin)]

Hi Joerg,

If I understand you correctly you wish to save your 2D gridded long/lat field
as a 1D array in NetCDF. The reason for this is to save space by compressing
the over-land grid points.

o The next version of Ferret will have a function RESHAPE(variable, new_grid)
that will handle this (already written and tested) -- the current Ferret
version does not have this.

o The decompression of your data (inserting missing value flags over land)
will, from the sound of it, require custom decompression routines. That is
probably beyond the scope of the current Ferret version's data management
capabilities. Again, the upcoming version will enable you to write your own
functions (FORTRAN, C, ...) into Ferret.

o An aside remark concerning styles and standards: NetCDF is designed to
promote self-describing data sets. The approach you are suggesting stretches
that notion pretty far. How about a simple IEEE stream for your data?

	- steve

==================================================================

On Apr 11,  3:25pm, Joerg Kaduk wrote:
> Subject: Data on land only
> Hi everybody,
> I'm really new to Ferret and so I've missed the solution to my problem
> in the docu or somewhere else... however after having leaved through
> whatever I could find I'm still somewhat unclear how to approach the
> following problem:
> I plan a to run a model on the land area on the globe only (say 1 deg.)
> resolution. This gives you about 15.000 gridelements on the underlying
> regular grid going from -180 to 180 E, -90 to 90 N. For storage reasons
> I want to save data only on the land points - on the ocean area I'don't
> do anything - in netCDF if possible. I would prefer to write the variables
> just as a one dimensional array - running through the land grid elements.
> And I'd like to use ferret to visualize the data. Now my question is
> can I tell ferret where to locate the data on the globe? I can store
> the latitudes and longitudes for the grid elements in a seperate file -
> preferably again just as a array running through the land points again in the
> same sequence as my other data.
> I think a more generic question would be: How do I tell ferret that the
> data in a file are on some unregular mask of a regular grid only? If I store
> a land mask alonag with the data in the netCDF file?
> Finally I could imagine the following trick could work somehow: I save my
> data running from south to north and west to east. Then in each band along
> longitudes there would be some ocean areas. For each ocean area I'd just
> save one missing value point and save in the netCDF file somewhere how
> long (longitudewise) my grid points are. This way one could make up a
> rectengular grid where some elements were very long ( the pacific is along
> the equator quite large...) How would I do this such that ferret would
> recognise this from the netCDF file without having to have a grid file
> spefified?  I'd really like to avoid having to specify some structure
> external to the netCDF files...
> Thanks for your patience and time everybody
> regards,
> Joerg
> --
> joerg@jasper.stanford.edu
>-- End of excerpt from Joerg Kaduk