NCEP Modeling Synergy Meeting Highlights: April 24, 2006


This meeting was led by Bill Bua and attended by Geoff Dimego, Joe Carr, Dave Michaud, Dave Plummer, Mary Hart, Jim Hoke, Zoltan Toth, Keith Brill, Mark Iredell, Keith Brill, Ed Danaher, Larry Burroughs, and Regina Nichols.  Additionally, SPC’s Steve Weiss and David Bright; and UCAR/COMET’s Stephen Jascourt attended by remote conference.


1. CCS

NAM-WRF:  Dave Michaud reported on the NAM-WRF parallel evaluation.  Information on the model has gone out via e-mail and other methods.  The field has given positive feedback regarding CCS meeting their needs for doing the WRF evaluation.  Scheduled implementation is still 12z 13 June 2006. 


North American Ensemble Forecast System (NAEFS): In the NCEP medium range ensemble forecast (MREF) system parallel there has been an overhaul of the file structure, with the new Ensemble Transform Kalman Filter (ETKF) used to create 14 perturbed runs plus a control, for 15 members per forecast cycle.  With the global ensemble file names changing, considerable coordination is required.  Brent Gordon or Dave Michaud should be informed if there are any problems with the file structure changes.  Implementation is scheduled for 12z 30 May 2006.  Note was made that there will be no combining of NCEP and Canadian Meteorological Center (CMC) ensembles yet, but that work is going on to make CMC ensembles look like NCEP post. For NAWIPS, all spaghetti and mean processing now running in HPC operations is being converted to run under production, so that for the next MREF parallel, comparison with the operational MREF will be easier to do.


Computer upgrade schedule: 

4/28/06:  Frost turned off.  New system in Gaithersburg will replace frost and snow over next several months.


7/15/06:  Delivery and installation of new Gaithersburg system starts (blue development machine replacement), with acceptance testing to be completed by 9/15/06.  Spin up production parallels for comparison with current operating computer.


July:  Fairmont, WV retrofit to get more power available for white (prod).  White down periodically during this time, blue will be operational.  Similar operational outage on 10/5/06 (~2 days), will also use blue for operations. 


9/15/06: Start conversion activities for blue and white. 


10/10-23: Load test of facility.  Old blue is still around during this time frame. 

Nov. 2006:  Fairmont ready in Nov., system installed for Dec acceptance testing, with likely Jan 2007 go by date.  Because of power generator shortage due to Katrina, there may be delays.  When new computers are live, blue and white will be disassembled, and NFS goes away. 


This will mark the most extensive upgrade we’ve ever gone through all at once. It was noted that at 10:30 a.m. every Tuesday, an IBM status meeting is held.




2a. Global Modeling Branch:

Mark Iredell reported that the next global implementation will be minor, and will not involve the new GSI data assimilation system, which is still undergoing modification and testing.  File name conventions will change January 2007 for the native “grid” files, because of the change to the native vertical coordinate from sigma to hybrid sigma-pressure.  A user guide will be issued in advance so that users will be able to change over without trouble. 


A new development real-time parallel using the hybrid sigma/pressure vertical coordinate will be started this week.  This GFS version was run with 45 cases from last summer, including hurricane cases, and from last winter.  Past cases are available for viewing.  Results look mildly positive, with the biggest impact on the East Asian jet, which holds together better because the terrain-following sigma coordinate is transitioned from right away.  (Note: in the parallel GFS, the transition from sigma to pressure starts at 2nd level, rather than at about 400-hPa as in the NAM-WRF.)  The time frame for implementation is 20-27 June; delays will postpone implementation because of the onset of hurricane season.


2b. Mesoscale Modeling Branch (MMB):

NMM/NAM-WRF:  Geoff Dimego reported that warm-season retrospective parallels have been redone a couple of times, correcting some things including changes to GSI and precipitation assimilation (at surface only).  The 2nd rerun was surprising; days 2.5-3.5 showed low bias in QPF, with drying 2-m dewpoints also.  Time will be spent to determine if a drift indeed exists, and whether it is different from the previous drift in the NAM-Eta.  MMB will continue to monitor this after implementation and will continue to work on problem.  Scores are competitive with NAM-Eta, so it appears the NAM-WRF will be good enough to be implemented. 


SREF:  Jun Du is moving SREF to 4 cycles per day, a 3rd quarter milestone.  The SREF group is finishing up work on the mechanics.  The SREF will become part of AWIPS build OB7 (but at 2x/day, not 4x).  Jun is also working on bias-correction of SREF and speeding up running of WRF members.  This will probably be implemented after computer switch next year (Jan. 2007). 


Real Time Mesoscale Analysis (RTMA):  Limited evaluation is currently being done of RTMA on 5-km NDFD grid from RUC-13 1st guess downscaled to 5-km resolution of the National Digital Forecast Database (NDFD). The RTMA includes 2D-VAR from the GSI for surface data, with anisotropic (terrain-following) covariances, and estimates of analysis error, which is not trivial for 2D/3DVAR.  Analysis evaluation is being done through cross-validation (i.e. Can withheld information be reproduced from the analysis?), which is being built into the RTMA.


DGEX: …will be continued for now with WRF instead of Eta.  ARW and NMM runs will be available when there is no hurricane run; and can be used in place of Fire Weather runs. Fire Weather runs cannot now be distributed to IMETS, so they may be suspended result to save computing time.


2c. Global Ensemble Prediction System: 

Updating of the global MREF is being done as discussed in NCO section above, including introduction of new products, and climate anomaly information for each ensemble member, for both NCEP and CMC. 


Future plans:  FNMOC (Navy) and JMA (Japan) are both interested in joining the system in the 18-24 month time frame.  We will be improving postprocessing in the next year.  In two-year timeframe, new kind of post processing will be done, based on a Bayesian statistical approach.  Both raw and calibrated data will eventually be made available.


2d. Short Range Ensemble Prediction System:

See MMB above.


2e. Marine Modeling and Analysis Branch (MMAB): 

The wave ensemble was implemented on 28 March 2006.  MMAB will be looking to implement the Great Lakes WaveWatch3 in Q4 of 2006, computer resources and transitions allowing (according to NCO, implementation would need to be done in July 2006).  GRIB product implementation includes the addition of wave steepness in May 2006 for AWIPS distribution.  Currently, this information is going to OPC (thru NAWIPS).




SPC:  Putting the SREF processing done by SPC at NCO has improved timeliness by 45 minutes than when it was done via SPC.  HPC:  WRF feedback and GFS ensemble review has, so far, not resulted in any “show-stoppers”. Regarding the GFS ensembles: ensemble mean highs and lows are stronger.  Regarding wave ensemble:  will 14 members be available when the GFS ensemble increases from 10 to 14 members?  It was also noted that the “Genesis tracker” for cyclonic system tracking, will also be affected.


4. The next meeting will be held Tues. May 30, 2006 at noon in room 209 with remote conference capability.