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

Re: Java GRIB2 decoder

On Tue, 16 May 2006, Rick Grubin wrote:

> Hello Robb,
>
> A quick question regarding the codetables you created for Unidata's
> Java GRIB2 decoder: did you find/have documentation on what various
> centers use for short names, or did you create your own?

Rick,

all the tables in my decoders came from the centers, i only modified the
units so they were udunits compatible. most of the short names are
acronyms from fortran programs so many are replaced with human
readable (meteorological) interpretations from CF convention as close as
possible. the purpose of the short names is to inform the user of the
meaning of the parameter not as a mechanism for matching parameters
between different type of datasets.

 It seems
> quite problematic to write code for GRIB2 decoding using codetables
> when there appears to be no agreement on short names in certain
> instances, which seems to defeat the purpose of a table-driven
> structure in the first place.  Thanks for any insight.

all the centers violate the table standards, thats why there are more than
one table. the only true interpretation is the numeric value for the
parameter and the numeric value may not be valid in another center's
table for the same parameter. the tables have versions that vary depending
on year. so it is possible for the same parameter to have a different name
depending on table version. the whole table setup is a can of worms. Paul
Hammer from FSL has been working on this for years.

the use of code tables in Java GRIB decoder is to interpret the data, not
to make comparison between dataset parameters from different centers. to
make comparisons would require a mapping of parameters from one center to
another center, that is beyond the scope of the program. but there is work
being done to create mapping between centers using CF, COORDS, GCMD, etc
conventions.

if you inspect the indecies that the Java GRIB decoder creates you will
see that the parameters are referred too by 3 numbers that represent
discipline, category, and parameter.

if you found inconsistencies,  i would like to know about them.

robb...


>
> -Rick.
> ----
> Rick Grubin      NCAR/CISL/SCD/VETS
> Visualization + Enabling Technologies
> address@hidden             303.497.1832
>

===============================================================================
Robb Kambic                                Unidata Program Center
Software Engineer III                      Univ. Corp for Atmospheric Research
address@hidden             WWW: http://www.unidata.ucar.edu/
===============================================================================