20020301: NEXRCOMP composite NEXRAD images

>From: Russ Dengel <address@hidden>
>Organization:  SSEC
>Keywords:  200203011521.g21FLFx21046 McIDAS NEXRAD ADDE

Russ,

re: archived NEXRAD data supported by the McIDAS NEXRAD servers

>     Thanks. I didn't know what level we were supporting.

Level III is composed of products.  Level II is the full volume scan
data.  The NEXR servers support the Level III stuff as that is what
is generally available.  Availability of Level II data  will change as
the CRAFT datastream becomes operational in the IDD.

re: what do you think about the composite images in NEXRCOMP

>      The ultimate in cool!  Way to go.

Glad you like these.  Please pass along the dataset info to interested
parties in SSEC (like Jerry R. and Rick K.).

>How do you generate them????

These are generated using GEMPAK.  Chiz wrote a routine that scans
directories for new NEXRAD N0R products whose "time" lies within the
window that he defines (right now at a half hour).  The newest image
from each radar that falls within the time window is included in the
composite.  The 1 km product is written out as a GINI image (my McIDAS
distribution includes a GINI server), and the 6 km national, 1 km
regional floater, and 10 km RCM images are written out as GRIB messages
that I convert to McIDAS AREAs.

We are including the 1 km national composite in our 'NIMAGE' IDD
datastream.  This stream contains all of the NOAAPORT channel 1 and two
products (GOES-East and GOES-West NOAAPORT GINI images).  It is getting
sent to only a small number of Unidata sites because of the data volume
and size of files (the VIS sectors are each 26 MB before compression
and come out 4 times an hour for both GOES-East and GOES-West).  The 1
km N0R composite that gets distributed in PNG compressed form is about
1 MB in size, but that mushrooms to 14 MB after uncompression.  Since
this product is created about every 8-10 minutes, you can see that you
can use up a lot of disk in a hurry.  We will be looking into adding
support for a compressed version of this product in both GEMPAK and
Unidata McIDAS (through my GINI server).  Back of the envelop
calculations told us that we should be able to get the file size down
to about 800 - 900 KB.  With this kind of size, it is entirely
reasonable to keep a couple of weeks of this product on line at a
time.

The 6 km national N0R, the 1 km regional N0R, and the 10 km RCM
composites are created from GRIB messages by a newly rolled
gribdec.tar.Z distribution that I am almost ready to release.  In
addition to the standalone GRIB to McIDAS GRID converter, gribdec.k, I
put together a grib2img.k application that allows one to go directly
from GRIB to McIDAS AREA (grib2img.pgm is a fusion of gribdec.pgm and
grdimg.pgm).  The RCM composites are available in NOAAPORT as GRIB
messages and XCD handles these with no problems.  The 6 km national and
1 km regional N0R composites are also generated by GEMPAK and output as
grib messages that are sent in the FNEXRAD feed of the IDD.  I simply
wrote an LDM action that converts these grib messages to McIDAS AREAs
upon receipt of the GRIBs and then I immediately inject the AREAs that
I generate from the gribs into the FNEXRAD feed after compressing them
with my AREA image compressor, area2png.  These products are reasonably
small in the broadcast, around 40 KB or so, and they only expand out to
300 - 400 KB.

The reason I include the RCM composite (which is also created by GEMPAK
back at the NWS) is so users can compare/contrast the RCM with the MDR
product and the Unidata composites.  The comparison is startling to say
the least!  The MDR product is simply horrible and it is misleading.
If you use it to look at the echos at the southern tip of Florida
today, you would never see that there are cloud streets down there.
The RCM product shows the features, but it is so coarse that it is
really unusable (and the values are VIP levels).  The 6 km national
composite shows the features well, but the 1 km product shows them in a
spectacular fashion.

Again, please tell Jerry and Rick about these images (and the BREF24
enhancement and stretch).  Oh, also, I had to modify BAR to be able
to work with INFO AUX blocks that contain more than 16 levels; the
change was trivial.  This mod will also be in my next McIDAS addendum.

Later...

Tom

>From address@hidden Fri Mar  1 14:47:15 2002
>Subject: Re: 20020301: NEXRCOMP composite NEXRAD images

Unidata Support wrote:

> Glad you like these.  Please pass along the dataset info to interested
> parties in SSEC (like Jerry R. and Rick K.).

I passed the info along to Rick and Jerry. I also sent it to Santek,
Achtor and Ackerman. It's always a good idea to show this stuff to
Achtor and Ackerman.  I think this kind of radar image makes more sense
than the individual NEXRAD site files.

re: How do you generate them????

I'm sure its a bit more complicated than just pasting the images
together. He must be doing some kind of filtering to get rid of ground
clutter, "best" value sensing,... It looks really clean. I do see some
radar edges along the Gulf coast, but that is to be expected.

Russ