Re: 20040625: GRIB problem in decoders-3.0.2 - bug?

• To: Arild Burud <Arild.Burud@xxxxxx>
• Subject: Re: 20040625: GRIB problem in decoders-3.0.2 - bug?
• From: Robb Kambic <rkambic@xxxxxxxxxxxxxxxx>
• Date: Mon, 28 Jun 2004 10:14:38 -0600 (MDT)

On Mon, 28 Jun 2004, Arild Burud wrote:

Robb Kambic wrote:
> Thanks again for the gribtocdl code. I wonder if version 3.0.0 had the
> same problem as 3.0.2?

Well, I didn't have the problem in the first version of
gribtocdl/gribtonc. Haven't tested the intermediate versions, but it
sort of made sense to see that valtime_offset was introduced together
with this problem.

> At first thought, I don't see how the
> forecasttime variable could cause this problem but maybe there is a
> dependancy that I over looked. I'll check it out.

Arild,

The dependancy that I over looked was the calculation of the variable
valtime_offset was delayed until gribtonc was invoked in the previous
version of gribtocdl. My calculation of valtime_offset in gribtocdl needed
to invoke the routine frcst_time() to obtain the correct value.

gribtocdl now creates the correct cdl for the sample data set you sent and
the results of gribtonc now only has one record.

As I said, changing the "Time 2 (P2) : 24" into "Time 2 (P2) : 0" in the
grib data completely removed the problem. At least I view it as a
problem when a "monthly mean" dataset becomes two datasets 24 hours
apart....
As "Time 2 (P2)" is mostly used as "Time interval between successive
analysis" it seemed logical that it could also be retreived as
"valtime_offset".

Anyway there is no hurry to fix this, but I just wanted to inform you of
my findings.

Thanks for the bug report, I put a version of gribtocdl.c in the ftp dir
if you want to try it out.  There will probably be a decoders release in a
couple of weeks.

Robb...

Arild
--
---------------------------------------------------------------------
Arild.Burud@xxxxxx       System Developer          Tel.: +47 22963035
http://www.met.no                                  Fax.: +47 22963050
Norwegian Service Centre for Climate Modelling,  http://noserc.met.no
Norwegian Meteorological Institute,    PB 43 Blindern,    N-0313 Oslo
---------------------------------------------------------------------

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