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

20020216: setting up non-decoding of certain grids in XCD (cont.)

>From: Unidata Support <address@hidden>
>Organization: UCAR/Unidata
>Keywords: 200202122315.g1CNFpx12009 McIDAS-XCD GRID

Tony,

I think that I have finally setup non-decoding on windfall the way you
requested.

Here is what I have attempted to keep:

MRF   all hours
AVN   0-24 hr
NGM   0-48 hr
ETA   0-48 hr
ECMWF all hours

You requested to keep the first 24 hour forecasts for ECMWF, but since
the real disk hot was the AVN, I didn't bother with it.

Now that disk space will no longer be the pressing issue, we should
continue the discussion of what you are really after gridwise.

Here are some things to ponder (some of these are repeats from an
earlier email; some are not):

1) AVN grids are delivered with several different "generating processes
   or model" (information taken from the GRIB Edition 1 document
   from NCEP):

   77 - 126 wave triangular, 28 layer Spectral model from "Aviation" run
   81 - Spectral Statistical Interpolation (SSI) analysis from "Aviation"
        run
   96 - Aviation Model

2) AVN grids come out in a several different projections:

   37-44 - "thinned" World MERC 289x145 (delivered in 8 octets)
   201   - Northern Hemispheric PS 65x65
   202   - National CONUS PS 65x43
   203   - National Alaska PS 54x39
   211   - Regional CONUS LAMB 93x65
   213   - National CONUS Double Resolution PS 129x85

   Good examples of the coverage of these various grids can be found
   in GEMPAK documentation that Chiz put together:

   201 - 
http://www.unidata.ucar.edu/packages/gempak/examples/models/grids/grid201.gif
   202 - 
http://www.unidata.ucar.edu/packages/gempak/examples/models/grids/grid202.gif
   203 - 
http://www.unidata.ucar.edu/packages/gempak/examples/models/grids/grid203.gif
   211 - 
http://www.unidata.ucar.edu/packages/gempak/examples/models/grids/grid211.gif
   213 - 
http://www.unidata.ucar.edu/packages/gempak/examples/models/grids/grid213.gif

3) the AVN model is run 4 times a day (not twice): 0Z, 6Z, 12Z, and 18Z

Right now, decoding on windfall is putting the 0Z and 6Z AVN grids into
one set of files and the 12Z and 18Z grids into another set of files.
The default setup for GRID decoding as sent with McIDAS-XCD was changed
some time ago to separate the different model runs into different
sets of files.  Since windfall is still running McIDAS-X 7.6, this
separation was never made.  The change should not be considered to
be drastic _if_ one uses ADDE access to the grids (i.e., if applications
using the gridded data are not looking to read specific GRID files).

The change to decode AVN runs into separate sets of files is done
in the file RTMODELS.CFG (located in uvaworkdata on windfall):

current RTMODELS.CFG on windfall:

SCRATCH=5001
NGM=  3 5051 120000 240000 480000
MRF=  1 5101 120000 240000 960000
MAPS= 0 5201 30000
MDR=  2 5301
AVN=  3 5401 120000 240000 720000
ETA=  1 5501 120000 240000 480000

recommended change sent out with Unidata McIDAS-X 7.7:

SCRATCH=5001
NGM=  3 5051 120000 240000 480000
MRF=  1 5101 120000 240000 960000
MAPS= 0 5201 30000
MDR=  2 5301
AVN=  3 5701  60000 240000 1200000
ETA=  3 5501  60000 240000 720000

The changed moved the GRID file number range from a start at 5401
(i.e., GRID5401) to 5701 (i.e., GRID5701).  The reason this was done
was there were no XCD-produced GRID files being written into the number
range after 5700.

(Also note that there was a change in the ETA setup as well.  This was
done to allow for the decoding into separate files of extended forecast
times in the ETA).

I would have changed the filing of the AVN grids for you (upgraded your
RTMODELS.CFG file), but I don't know how you are doing your processing
(ADDE use of dataset group/descriptor.position or simply read from a
particular GRID file).

It would seem that you probably need one or two of these grids for your
processing.

Looking forward, it would seem that you might benefit from the CONDUIT
grids that are on grid 3 (65160 points, 181x360, 1 degree, MERC).  I
just got decoding of these working with XCD the other day.

Given your finite amount of disk space, we should work out exactly
which grids you will benefit from having and which are expendable.  If
you get to the point of wanting to ingest and decode the CONDUIT, you
will need to upgrade your McIDAS installation.  I realize that this
will be a big move (given your locally developed code), but I can help
in the transition.

Talk to you later...

Tom