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

20010419: gdradr

Scott and Steve,

I uploaded a tarfile of my GDRADR tree to the hp1 upc account.
The tar file includes the ancillary pdf, prm, txt and err files
as well. The Makefile, while not useful with build, will show the libraries
used in linking.

I have a running web page which shows some GDRADR products at:
http://motherlode.ucar.edu/unidata/images/nids/gdradr.html
The national and regional mosaics are plotted from the grid files using gdplot2.
The precipitation type composite uses RUC fields interpolated to the same
grid as the radar. Interesting sites are determined by using the GWFS() function
in gdgsfc back to the nexrad.tbl locations and then using sflist to pick out
the top 5 weighted reflectivity values. If interested, I can provide you with
the scripts.

The GDRADR program will look for a NEXRIII entry in datatype.tbl for the
naming convention used for NEXRAD level 3 products. 
The datatype.tbl entry I am using is:
! Unidata addition for GDRADR
NEXRIII  $RAD/NIDS/%SITE%/%PROD%  %PROD%_YYYYMMDD_HHNN  CAT_NIL SCAT_NIL 10 
2880 -1

If NEXRIII isn't found in the datatype.tbl file, then the program assumes that
the NEXRAD files are named with:
path: $RAD/NIDS/%SITE%/%PROD%
filename: %PROD%_YYYYMMDD_HHNN

The %SITE% template will be replaced by each station entry found in the 
STNFIL (nexrad.tbl). The %PROD% template will be replaced by the GFUNC parameter
(N0R, NTP, BREF1, etc). I use gfunc=n0r for the images on the web pages above.

Currently, the first time through the station list, the gqbnds() of each radar
must be computed in the grid domain (this is slow ~10 minutes for
360,000 grid points and all nexrad.tbl stations). These values are saved so
that the process is much faster, ~70 seconds on all subsequent times.
I am looking at ways to speed this up, either with a table of previously
computed bounds for a know grid projection, or by eliminating the gqbnds() call
for a more efficient way of quickly/trivially rejecting sites not in the domain.

I am working on an alternative output instead of grid. At this point, I have
GINI format output. The projection limitations with using the GINI or NIDS
formats for floater mosaics is somewhat of a hassle.

I have also fiddled with an addition to gdplot2 adding a TYPE for just filling
in the grid box rather than using color filled contours. This works good and is
faster than TYPE=f. The results are generally acceptable for the size grid boxes
I am using. 

Let me know if you have questions, problems, suggestions etc.

Steve
------------------------------------------------------------------------------

Steve Chiswell
Unidata User Support