15km FIM global model runs out to 10-day duration, output every 3h
Twice daily forecasts run on TACC (Texas Area
Computing Center) supercomputer as part of NOAA-TACC demonstration
for improved hurricane forecasts.
GFS analysis (spectral native) used for FIM initial conditions,
more information at
fim.noaa.gov
14 Oct 2008
Real-time runs are finally back as of yesterday!
Thanks to Chris Harrop, once again!
8 Oct 2008
No runs since 1 October at TACC. Ranger was down for 5-day
maintenance until late Monday 6 October. Executables
all have to be rebuilt since new MPI-related libraries
are now required -- this is now underway.
22 Sept 2008
Runs continuing twice daily, successful output for Gustav, Hanna, Ike, etc.
The 3 Sept 00z FIM TACC G9 (15km) run
did not complete, was resubmitted this morning,
and will hopefully complete this evening.
Tracks for tropical storms identified by NHC
have now been added to the 10-m wind FIM forecast graphics
for the western Atlantic area
starting with the 0312z runs. Take a look at
10-m winds under
http://fim.noaa.gov/fim244
.
Tracker data now available for ESRL G8-30km FIM run.
Last night's 00z 1 Sept run available at
http://fim.noaa.gov/G8-tracker
.
These runs are also experimental, run on
smaller ESRL wJet supercomputer
instead of TACC Ranger.
FIM TACC G9 (15km) runs running for todays 12z run (3012),
will send out update when some hand-moved plots
are available, may be tomorrow.
Sat 30 Aug 2008 22z
FIM TACC G9 (15km) runs for 2900, 2912, and 3000 all ran.
The data is getting back to GSD, but not to GSD's web pages
or to Brent Gordon's ftp web site yet.
There are QuickTime movies of the 10-m wind output from the
last three FIM-TACC-15km experimental runs at
http://fim.noaa.gov/movie
.
(Thanks to Chris Harrop, who has been working on
Ranger today.)
These files at 8-9 MB, a bit on the big side.
Also, we have to move them in by hand, but will keep doing
so for the rest of the weekend.
There are also tracker output files available for the
same 3 runs at
http://fim.noaa.gov/tracker
.
Same deal - being moved by hand, NOT automatically.
No SLP or wind max data yet, only position.
Fri 29 Aug 2008 23z
12z run from yesterday (2812) complete, with all graphics became available
this morning at
http://fim.noaa.gov/fim244TACC
. This is the first "credible" 15km FIM yet, with all
known problems solved (soil moisture problem, latent heating problem)
00z (2900) completed this morning, but /scratch file system
at TACC (where all FIM processing is donw)
went down during post-processing.
/scratch at TACC came back up after 21z today but back down at 2240z.
Scripts left in place to attempt to do the twice-daily FIM runs
Update from Chris Harrop at 1640z today
An automated hopscotch data transfer of TCVitals from GSD to TACC is set up and working.
Tracker is fully automated at GSD (no track plots yet, and winds still missing).
Tracker is fully automated at TACC (no track plots yet, and winds still missing).
Side benefit of 1 and 2 is that the production of the small grib files are now automated as well.
Automated transfer of results from TACC to GSD is implemented and should take effect this afternoon.
Included in this transfer are: graphics, tracker output, and the small grib files
3 speed-up changes are in the works for FIM for next week:
increased halo size, post-processing of output times as available,
parallel version of FIM pre-processing. Total speed-up might
be as much as 3 hours.
Update on ensembles: Code ready to go at TACC to run FIM ensembles
but that won't be tried until the single determininstic FIM is working automatically.
The first test will be for a coarse 240km (G5) grid next week hopefully.
Fri 29 Aug 2008 15z
Back-door ftp from NCEP CCS/dew to GSD set up
for GEFS ensemble initial conditions.
These files are now getting to GSD every 12h. (Kevin Brundage)
Fri 29 Aug 2008 03z
Tracker running for GSD FIM now running automatically
but no automated graphics yet. Update today from Chris Harrop.
Thurs 28 Aug 2008 23z
TACC experimental FIM runs
Twice daily runs are cron'd to run automatically.
15km FIM runs out to 10 day duration, output every 3h
GFS analysis (spectral native) used for FIM initial conditions,
more information at
fim.noaa.gov
Tracker code run experimentally at GSD from FIM output,
not likely to run automatically from TACC-FIM for this weekend but
work will continue tomorrow toward this goal.
Web products from experimental TACC-FIM are also not yet automated.
Currently, different GSD programmers are ftp'ing these
products by hand to GSD and getting them to the right location
by hand.
GSD is also working toward producing the small GRIB output file
from TACC-FIM runs (12 fields, including those necessary
for the tracker). This also is in testing, but not
yet automated.
GSD is also working on 3 different changes to speed up the
processing of the FIM model, which is currently 8.5-10h
for each 10-day forecast. All of these changes will be
held until next week.
If there is some processing problem on TACC over the 3-day
weekend, GSD is not able to immediately fix it and will
not have programmers following it through the weekend.
Meteorological value of first 4 TACC-FIM runs
The first near-real-time run - 26 Aug 12z (2612) ran
with plausible results, even if not objectively verified.
The next 3 runs (2700, 2712, 2800) had a obvious problem
(killing all tropical storms!),
which was found to be a lack of latent heating from precipitation.
This was a result of change to a FIM parameter to speed up
runs.
The 2812 run has the latent heating problem solved,
and also includes a fix to a separate problem regarding soil moisture
in all previous runs.
GSD FIM runs
These continue twice daily at 30km resolution out to 7 day duration.