Re: [ferret_users] seeing some strange output in curvilinear grids

[Ansley Manke <ansley.b.manke@xxxxxxxx>]

Hi,
The CF standard for NetCDF datasets does not allow missing data in 
coordinate variables. Ferret should issue an error message when there is 
missing data in the longitude or latitude coordinate variables for the 
3-argument plot commands. So in general, we'd expect to have seen this:

yes? fill variable, longitude, latitude
 **ERROR: value out of legal range: X coordinates
          missing value where val is needed

Your dataset is being read in such a way that the missing-value flag is 
not used quite correctly at the point within Ferret where that is being 
tested. Ferret should be able to have handled it, so there is a bug, and 
I see what its fix will be.

As an aside, a couple of comments about your dataset: The hdf-5 dataset 
is accessed by hdf5 through the netcdf-4 library, and so we can read the 
data in Ferret.  It's not quite conforming to the CF standard. For 
instance, when you open the dataset, you'll see

yes? use satellite_missing_xcoords_ycoords.h5
 *** NOTE: Internal err: unknown data type

The data type of the variable IMG_WV is not coming through the netCDF 
calls as of the recognized data types SHORT, INT, FLOAT, etc.  Overall 
things seem to work fine, but there may be details to deal with; for 
instance if you wanted to write a subset of the data using Ferret, you'd 
need to set the output type

yes?  save/outtype=int/file=subset.nc  img_wv[i=1:400,j=1:400]


Ansley

On 2/21/2014 5:52 AM, Ghansham Sangar wrote:
> Hi All
>
> Kindly see this one.
> I am trying to plot a curvilinear dataset
> This is an hdf5 product that follows Netcdf-4 classic data model
> and CF conventions.
> I am able to open it.
> But there are a few issues.
> Kindly check attached files.
>
> regards
> ghansham