[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: GRIB to netCDF

Hi Tom,

> We would like to translate our GRIB format files to netCDF.  What would you
> recommend?

Well, I don't currently know of any generic GRIB to netCDF translators.  We
have one (gribtonc) that is fairly specific to some of the GRIB products
from NMC.  It writes many GRIB products into a single netCDF file, using
knowledge of the grids on which the GRIB products are defined to "stitch"
subgrids together into a global grid and to stack up 2-D grids at different
layers and forecast times into 4-dimensional arrays of parameters such as:

        float   Z(record, level, y, x) ;
                Z:long_name = "geopotential height" ;
                Z:units = "gp m" ;
                Z:valid_range = -1500.0f, 50000.0f;
                Z:_FillValue = -9999.f ;
                Z:navigation_dim = "nav" ;

where "record" is determined by the reference time and forecast time of a
model run.  gribtonc is limited in the GRIB features it supports (it doesn't
handle the Grid Description Section or second-order packing) and it requires
that GRIB products be "wrapped" in a WMO envelope.  gribtonc has some other
limitations:

   - inability to decode into more than one output file;
   - inability to handle recently added GRIB products, including
     "quasi-regular" thinned grids;
   - inability to decode GRIB data directly into ASCII form;
   - inability to use any kind of packing in the netCDF data: only arrays of
     floats are used.

I'm currently in the process of writing a new GRIB to netCDF decoder
(gribdec) that will handle the Grid Description Section, improve
portability, permit more flexible specification of how multiple GRIB
products are to be mapped into netCDF files, and handle a stream of raw GRIB
products with no WMO envelopes.  I hope to have a beta test version
available soon.  The purpose of the new decoder is still to handle GRIB
products we get from model output runs at NMC and ECMWF, but it may also be
more generally useful.

There are lots of issues in how GRIB information should be represented in a
netCDF file, for example how to represent georeferencing information in the
Grid Description Section in an understandable way that application programs
can use.  For an example of how we have tentatively decided to represent
georeferencing information with a single netCDF dimension and a set of
variables that use that dimension, see

   ftp://ftp.unidata.ucar.edu/pub/netcdf/Conventions/NUWG/ruc.cdl

______________________________________________________________________________

Russ Rew                                                UCAR Unidata Program
address@hidden                                          P.O. Box 3000
http://www.unidata.ucar.edu/                          Boulder, CO 80307-3000
______________________________________________________________________________