[Met_help] PrepBufr files

[Lacey Holland]

Hi, Brian!


There's a very helpful website where I've been getting archived prepbufr 
data.
It is:  http://nomads.ncdc.noaa.gov/


 From there, I'm able to access GFS prepbufr files that have been 
archived.  There are a few dates missing from that site, but all-in-all 
it is pretty good.  I think the NOMADS website has prepbufr files for 
6-h intervals for GFS/GDAS.  I'm not sure about prepbufr files for other 
models and their frequency of creation, though.


The time of observations can, indeed, be determined in prepbufr files.  
You should be able to break the prepbufr files up into smaller chunks 
using the "beg_ds" and "end_ds" flags in the PB2NC configuration file, 
and then output each to NetCDF files.  The default is to search for data 
within a +/-5400 second (1.5 hour) window for the time listed for the 
prepbufr file.  You could change those flags to search for data within a 
series of 1-h time windows, centered around the time of your forecast's 
valid time.  The "beg_ds" and "end_ds" designate time (in seconds) 
relative to the time listed for the prepbufr file.


For the GFS files output in 6-h intervals, I would create six different 
configuration files for pb2nc, all the same, except varying beg_ds and 
end_ds.  I might change them to something like this:

1) beg_ds = -10800
     end_ds = -7200

2) beg_ds = -7200
    end_ds = -3600

3) beg_ds = -3600
     end_ds=  0

4) beg_ds = 0
    end_ds = 3600

5) beg_ds = 3600
    end_ds = 7200

6) beg_ds = 7200
    end_ds = 10800


... or something to that effect.
 
If you want them centered in a different way (I've only broken them into 
1-h chunks centered on the half-hour), you could modify beg_ds and 
end_ds to suit those needs (for example, if you want to verify at 0500Z, 
you may want data between 0430Z and 0530Z, or even 0445Z and 0515Z).  
Just calculate how long in seconds those occur relative to the time 
listed for the prepbufr file, then change beg_ds/end_ds accordingly.


In the above example, the observations occurring exactly on the hour 
(+/-10800, +/-7200, etc.) are counted twice, so you should also decide 
how you want to adjust your times to prevent any observation from being 
counted twice. 


Another thing you might note is that currently, we can't combine 
prepbufr files, so if you would like to combine data from two different 
prepbufr files, we don't currently have that capability.  However, after 
consulting my colleague John, it sounds as though that could be done 
with the NetCDF files that are produced, but right now that is up to the 
user.  We are talking about adding in a way of handling that in future 
releases, as that might be really important for some users!


Anyhow, I hope that answers your question.  Let me know if there's 
anything else I can do to help!


--Lacey Holland




Brian Gaudet wrote:
>
>
>
> Hello,
>
>
> I had a couple of questions regarding PrepBufr files:
>
>
> *)  Do you know of a place where PrepBufr files are
> being archived (going back about 1 year)?
> Specifically those of the same format as at the
> NOMADS site.
>
>
> *)  Is 3-h the maximum frequency of PrepBufr file
> output, or is 1-h available somewhere?  Within
> each file, can the time of observations be
> determined to the nearest hour?
>
>
>
> Thank you for any information you can provide.
>
>
> Sincerely,
>
>
> Brian Gaudet
> bjg20 at psu.edu
>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Met_help mailing list
> Met_help at mailman.ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/met_help
>   


-- 
____________________________
Lacey Holland

UCAR/NCAR, Research Applications Laboratory
P.O. Box 3000
Boulder, CO  80307
(303) 497-8396