Angel,

I have played around with your file today. The easiest way to get it into
OpenDX was to add the netCDF attribute

  sh:field =  "Spec.Humid., scalar, series" ;

The field attribute is necessary to tell OpenDX what variables are of interest
to import. Also, since the data is a time series it is necessary to explicitly
state this in an attribute. Otherwise the variable is imported as a 4D matrix
and needs to be sliced before it can be rendered.

I tried to use ncdump/ncgen to insert this info, but failed, so in the end I
had to write a program to add this info in the netCDF file. The file is
available at http://www.pdc.kth.se/~smeds/Angel/modify_netcdf-1.c

This will import the data without the coordinate positions. If you want the
coordinate positions as stated in the file the file needs have the attributes

>               sh:field = "Spec.Humid., scalar, series" ;
>               sh:positions = "zpos, product; ypos, product, compact; xpos, 
> product, compact" ;

and the following variables need to be added

>       float xpos(TWO, THREE) ;
>       float ypos(TWO, THREE) ;
>       float zpos(z, THREE) ;

This is accomplished by the program available at
http://www.pdc.kth.se/~smeds/Angel/modify_netcdf-2.c

However, as the Dx/Dy increments are 5000 while the increment in the z
direction is 1/100 of this, the model gets not very nice looking. It gets
very flattened out and thin.

The programs have been written using netcdf version 3, if you are using
version 2 there will need to be minor modifications made to the code.

The way the data is layed out in the file and the way data is assumed to
be in the netCDF importer makes the image to have the original x and z
axis swapped in the display. This could be fixed easily with a
transformation module.

I also added a compute to the generated net to map all the out-of-place
numbers to 0.0. It makes the plot looks nicer, but may not be correct
for your application. The net is available at
http://www.pdc.kth.se/~smeds/Angel/Angel.net

Good luck with your work on the data!

/Nils Smeds
-- 
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
   Nils Smeds                        http://www.pdc.kth.se/
   Center for Parallel Computers     e-mail: [EMAIL PROTECTED]
   Royal Institute of Technology     Voice:  +46-8-7909115
   KTH                               Fax:    +46-8-247784 
   S-100 44 Stockholm, Sweden        Office: OB2, room 1546
-----------------------------------------------------------------------

Reply via email to