EMC/HPC/MPC/NCO/CPC synergy meeting highlights 04/30/01
This meeting was led by Stephen Jascourt and followed a generic agenda denoted by the subtitles below. Attendees included Geoff DiMego, John Ward, Bob Grumbine, Bill Bua, Keith Brill, Ed Danaher, Kevin McCarthy, Zoltan Toth, Pete Caplan, James Partain, Naomi Surgi, Dave Reynolds, and Ralph Petersen.
1. IBM SP
John Ward reported that the B side is not fully operational. Testing needs to be done first on individual model runs and then using the B side for generating and shipping operational products for one day. B side will be only half available for development on the day of this testing, but A side operational part will still be running the operational suite in case the test encounters unexpected problems (so products still get out). The OSO server will be set up to grab from whereever it finds products (A side or B side). Thereafter, preventive maintenance never shuts down the system or creates delays, as operations can continue whichever side is not being worked on.
Firewall testing on B side will be conducted Thursday May 3, so be prepared for potential outside access problems.
Delays in implementation of Eta extension to 84 hours caused by peculiar IBM system problem which has been side-stepped (so model runs ok now) but not solved (may affect future implementations).
NTRANS has problem that it cannot make graphics until last model output time available. Since Eta runs 0-60 hours early, then 60-84 hours later, NTRANS Eta graphics are delayed. So NTRANS job will be done twice, first with only hours 0-60. Apparently, modifying GEMPAK to append later times onto NTRANS loop (so tack on 60-84 hours when it becomes available) is not on the list of priority changes for Mary Desjardin's group.
Parallel EtaX is now available to HPC in NAWIPS.
May 15: Global model upgrade (all systems - AVN, MRF, ensembles)
May 22: GFDL hurricane model coupled with Princeton Ocean Model (Atlantic basin only)
May 29: Short range ensembles ("Realtime Test and Evaluation", not full implementation)
mid June: Eta spring bundle (3D-VAR, precipitation assimilation, and land surface upgrade)
Ralph Petersen explained "Real-time Test and Evaluation":
2. Notes from EMC
a. Mesoscale Modeling Group: Geoff DiMego reported that the spring upgrade package will likely go to CAFTI on May 24 for a mid-late June implementation. One of the three components, precipitation assimilation, has been in the EtaX full-resolution parallel, while the other two components, new 3D-VAR and land surface package, are being tested individually in 48-km parallels and will go into the bundled full-resolution EtaX test in roughly two weeks. Alaska Region has volunteered to look at performance of the land surface package, which includes more realistic handling of snow (such as allowing patchy snow so ground can warm above freezing).
The Eta "threats" runs are in "Real-time Test and Evaluation" mode (see above). Gempak files are being made for NAWIPS on a 10-km grid. Grib files for a limited product suite (including surface and boundary layer fields and QPF) are now on the NCEP server. These grids need to also get on the OSO server for easier access by regions having a direct link to that server. For instance, Western Region Headquarters distributes grids from the OSO server to local offices which bring them into AWIPS.
A 10-km nest over Hawaii will also be run in "Real-time Test and Evaluation" mode as additional guidance at 00 UTC and 12 UTC. Rather than beiing replaced by an Eta nest, the RSM will continue running until Pacific regions says they don't need it. Also, Puerto Rico will get a tiny nest at 06 UTC and 18 UTC.
b. Global Modeling Group: Naomi Surgi reported good news that the new AVN (with cumulus momentum mixing and explicit cloud water) produced improved GFDL hurricane forecasts. The GFDL atmospheric model has several upgrades including improved boundary layer physics. It will be coupled with the Princeton Ocean Model over the Atlantic basin - but not over the Carribean and Pacific basins (ocean models not yet configured for them). Also, as has been reported previously, the new AVN greatly reduces spurious tropical vortices. Naomi Surgi thought this may be due to the parameterized cumulus momentum transport causing mixing out of shallow low-level westerly flow which develops in low-level model vortices.
Zoltan Toth reported that the new AVN/MRF physics will also be implemented in the global ensembles. Pete Caplan reported that the model upgrade does not include use of half-degree SST data (the Real-Time Global 2D-VAR analysis now used in the Eta).
c. Short range ensembles (SREF). Ralph Petersen, together with Geoff DiMego and John Ward, reported on the short range ensembles. They will be run between 0-1 UTC and 12-13 UTC with 09 UTC and 21 UTC initial conditions. This is the only available time slot in the IBM SP production schedule and it allows output to be ready in time for use with the 00 UTC and 12 UTC Eta package. Timings reveal that the RSM does not optimize as well as the Eta, taking almost 3 times as long to run, and further optimization would take considerable IBM staff time that could be used for work on other model codes.
"Real-time Test and Evaluation" mode will commence in two phases: starting May 29 output will be available on the web as they are now. The new status will bring increased reliability and timeliness. Starting July 2, the ensemble combined products (spaghetti diagrams, etc) will be available as grib files on the NCEP server and the service centers will be able to generate NAWIPS files. Also thinned output of individual ensemble members will be available in grib format, as they are now for the global ensembles.
NCEP service centers will be able to request ensemble products for centralized production on the IBM SP if calculation of those parameters requires data not output in the grib files for ensemble members.
Full operational implementation including dissemination of some products to AWIPS is projected after around a year, during which preparations are needed in training forecasters on use of short-range ensembles, calibration of ensemble probabilities for ensemble guidance products, and readying of AWIPS for this additional data.
d. Ocean Modeling Branch (OMB): Bob Grumbine, who happened to be sitting in on the meeting to observe, noted that the OMB has marine satellite winds operational now with more on the way (some satellites/algorithms available presently, others in various stages of development) and is heading toward 1/8-degree SST analysis and sea ice analysis, and in the more distant future, a sea-ice model.
A major upgrade of the RUC model is scheduled for June 30 as per Stan Benjamin's Notice of Intent to Change. However, this is viewed as an unrealistic target.
3. Input to EMC from Operational Centers
One internal-to-NCEP matter was raised which would enable improved ability of HPC to provide freezing levels to the RFCs for their hydrological models.
4. Next Meeting Proposed Monday June 4, 2001 at noon in room 209. Note that the regularly scheduled meeting date would fall on the Federal holiday of Memorial Day. Delaying the meeting for one week will put it after the three implementations in May so that any related unexpected issues can be discussed.