EMC/HPC/MPC/NCO/CPC Synergy Meeting Highlights 12/01/03
This meeting was led by Peter Manousos and was attended by Geoff DiMego, Jim Hoke, Hua Lu Pan, Dave Michaud, John Ward, Keith Brill, Bill Bua, Stephen Jascourt, Ed Danaher, Joe Sienkiewicz, and Larry Burroughs.
1. IBM SP and CCS
John Ward reported that the CCS continues to run smoothly. An upgrade to the system is planned for June 2004 (3x or more faster than the current system with 2x memory per node). Details of the new systems configuration has still yet to be negotiated with the vendor.
2. Notes from EMC
a. Global Modeling Group: Hua-Lu Pan reported the reason that the GFS failed for two cycles the last week of November was that the vertical advection speed hit the CFL limit in the vicinity of the Himalayas. A crisis change will be implemented to fix this problem before Christmas. The GFSx (parallel GFS including sub-grid scale blocking effects of orography) was started on 22 November 2002 and is expected to have a significant impact on lee-side cyclogenesis. So far there has been only a minor impact observed. The NCEP service centers will be asked to examine the output during the December-January period and give their feedback.
b. Mesoscale Modeling Group: Geoff DiMego reported that the fall bundle now should be implemented in January 2004. However, preliminary evaluation is leading to the decision to NOT include the radiation package in the January implementation. Otherwise there have been no significant changes for the bundle details or the Eta and WRF roadmap outlined in the September 30, 2003 Synergy Meeting Highlights. Additionally, the Eta will be run out to eight days over a smaller domain as a temporary method for downscaling the GFS forecast for use in IFPS, though communications issues for distribution to AWIPS remain unresolved. If the change is implemented, the Eta 0-84 hour forecast will finish earlier - roughly 25 minutes earlier for 60 hours, an hour earlier for 84 hours. The 8-day extension would be accomplished using the same amount of computer resources as the old 60-84 hour forecast, by reducing the domain for the extension to the CONUS and nearby areas only. However, the computer resource used for the extension to eight days will be given to the GDAS when the GDAS needs more resources, while the earlier Eta output schedule will continue.
c. Global Ensemble Prediction System: Hua-Lu reported for Zoltan Toth that adding a control and 5 pairs at off-time runs (06 and 18 UTC) runs is running in parallel now and still on track to be implemented before the Holiday Moratorium.
d. Short Range Ensemble Forecast (SREF) Group: The diversity package of the SREF (32km horizontal resolution) is currently running in parallel and now has a target implementation date of Jan 20, 2004 for two times per day runs. There are plans to run it four times per day by mid- to late 2004 to complete the symmetry of running all models in the production suite four times daily. Details of this package are found in the September 30, 2003 Synergy Meeting Highlights.
e. Marine Modeling and Analysis Branch (MMAB): Larry Burroughs reported that the Global visibility fields, Eta visibility fields, and Global Vessel Icing are in parallel. The response from users has been positive, but an operational implementation date has not been settled upon just yet. Discussion with AWIPS Program Office and TOC are ongoing and focused on getting neural-net-filtered SSMI data on AWIPS by mid 2004 for AWIPS build OB4.
HPC inquired to NCO on accessing full grids of the fire weather (8km NMM) runs. NCO and EMC stated that these may be accessible internally to the NCEP centers before Christmas. HPC also inquired to EMC on the possibility of accessing Eta KF runs on a routine basis. EMC responded that the new diversity package of the SREF will have at least one member running the Eta using a KF scheme and at a higher resolution (32km) than the 48km version HPC typically views. Finally, HPC inquired if EMC could make the height of the wet bulb freezing level available in the operational post processed output. EMC responded that the Meso group is already working on the code with a target of implementation either before the holiday moratorium or shortly thereafter. The Global group has suggested that code be developed by HPC and then “JIF’d” to run on the CCS.
The Meso group inquired to HPC about both the technical aspects and HPC perception of running the Workstation Eta at HPC. HPC responded that it runs the Workstation Eta on a Linux box with a limited domain (does not cover the entire US) at 32km resolution out to 30 hours. The domain is movable each cycle via a TCL/TK interface by the forecaster to focus on the “event of the cycle”. The Workstation Eta at HPC is also run using the GDAS with either a KF or BMJ parameterization (selectable each cycle by the forecaster). This was found to exhibit inferior performance to the 32km Eta run with a GDAS by EMC on the IBM SP during LAST winter season. There is some advantage to running the HPC Workstation version of the Eta during convective season, but since the Eta is already running at 12km resolution operationally, and the time the output of the Workstation Eta is available lags behind the operational model by at least 5 hours, HPC concludes there is little advantage of running the Workstation version of the Eta. Therefore, interest has shifted at HPC to utilizing and testing the performance of the 8-km NMM (now WRF-NMM) fire weather nests mentioned in the previous paragraph. Additionally, HPC runs the Workstation Eta over South America out to 84 hours at 32km resolution. Further, HPC has recently begun running a 48km resolution version covering the entire CONUS once a day (0000 UTC) out to 180 hours.
4. Next Meeting Monday January 26th, 2004 at noon in room 209.