[Met_help] Problem in GDAS PrepBufr dataset

[John Halley Gotway]

NOMADS Support,

I reported a problem to you on 2/13/2009 about the GDAS PrepBufr observations files that are posted on the NOMADS data server.  I've listed that original message below.

The problem regarding the 18Z files being shifted by one day continues to exist on your server.  For my current work, we're using data between 20080601 and 20090603, and the problem exists for all
those dates.

In addition, I've found a few other files for which the timestamp is incorrect:

(1) http://nomads.ncdc.noaa.gov/data/gdas/200807/20080708/gdas1.t18z.prepbufr.nr, the timestamp is wrong.
(2) http://nomads.ncdc.noaa.gov/data/gdas/200807/20080709/gdas1.t00z.prepbufr.nr actually contains data for 00Z on 20080708.
(3) http://nomads.ncdc.noaa.gov/data/gdas/200807/20080709/gdas1.t06z.prepbufr.nr actually contains data for 06Z on 20080708.
(4) http://nomads.ncdc.noaa.gov/data/gdas/200807/20080709/gdas1.t12z.prepbufr.nr actually contains data for 12Z on 20080708.

Please let me know if you'll be able to resolve these issues.  As I mentioned in my original email, when I check the GDAS PrepBufr files on the NCEP HPSS directly, the data there IS correct.  So
somewhere along the line, the timestamps for the 18Z files are getting shifted.

Thanks,
John Halley Gotway
johnhg at ucar.edu

John Halley Gotway wrote:
> NOMADS Support,
> 
> We've found what appears to be a problem in the GDAS PrepBufr datasets available on the NOMADS data server:
> http://nomads.ncdc.noaa.gov/data/gdas/
> 
> For each day, there are 4 PrepBufr data files available in the path: http://nomads.ncdc.noaa.gov/data/gdas/YYYYMM/YYYYMMDD/
> (1) gdas1.t00z.prepbufr.nr should contain obs from 21Z the previous day to 03Z the current day.
> (2) gdas1.t06z.prepbufr.nr should contain obs from 03Z to 09Z for the current day.
> (3) gdas1.t12z.prepbufr.nr should contain obs from 09Z to 15Z for the current day.
> (4) gdas1.t18z.prepbufr.nr should contain obs from 15Z to 21Z for the current day.
> 
> However, we've found that the timestamps of the observations in the 18Z are consistently wrong.  The timestamps are for the prior day.  For example, you'll find that on 20090131, the
> gdas1.t18z.prepbufr.nr file contains observations between 15Z and 21Z for 20090130... not 20090131 as it should.  However, the 00Z, 06Z, and 12Z files do contain the expected timestamps.
> 
> Now, I don't know if the 18Z files contain the wrong observations or if it's the timestamps that are wrong.
> 
> I retrieved an 18Z PrepBufr file for the same date directly from NCEP's HPSS file system: gdas1.t18z.prepbufr.nr for 20090131
> (It's archived on the HPSS in the file /hpssprod/runhistory/rh2009/200901/20090131/com_gfs_prod_gdas.2009013118.tar)
> I found that this file is fine.  It does NOT have this same problem.  It contains observations from 15Z to 21Z for 20090131.
> 
> It appears that this problem has been around a while.  The person who originally found this issue, saw it in PrepBufr files from June through August, 2008.
> 
> Thanks,
> John Halley Gotway
> johnhg at ucar.edu
> 
>