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

20010313: dcgrib2 radar mosaic flushing problem?

>From: "Arthur A. Person" <address@hidden>
>Organization: UCAR/Unidata
>Keywords: 200103131521.f2DFLUL21846

>Hi...
>
>We're trying to use the radar mosaic's decoded by dcgrib2 as soon as they
>become available and have noticed that often (maybe half the time) the
>latest hour's stuff doesn't become available until close to the next hour
>when the new data comes in.  Could there be some sort of buffer flushing
>problem in the gempak output for these data?
>
>                                      Thanks.
>
>                                        Art.
>
>Arthur A. Person
>Research Assistant, System Administrator
>Penn State Department of Meteorology
>email:  address@hidden, phone:  814-863-1563
>


Art,

The GEMPAK write routine does flush the output (writing all its buffer out),
but there is no "FLUSH" in a fortran write statement, so the output in this
case is being held up until the file is closed, or another write to that
file occurs. Since the data only comes in at 30 minute intervals, you see the 
header information updated in the GEMPAK grid file, but can't access the grid
data for some time (particularly troublesome when trying to use LAST for
gdattim). Since dcgrib2 has 5 output streams open, it will only close
down a stream if it is needed for another data set- so during the day when 
other models aren't coming in, this is most problematic. 

The only current solution for this problem is to create a separate 
PIPE -close action for this data set. I am looking at creating a timer
in dcgrib2 that will force a close of the output file if no data has been 
written to that stream recently. I am also working on a NEXRAD level III
mosaic program that will produce better resolution mosaics than the
RCM that comes on NOAAPORT:
http://motherlode.ucar.edu/unidata/images/nids/gdradr.html

Sorry that this is causing you trouble- I hope to have some solutions available 
in 5.6.B in a couple of weeks.

Steve Chiswell