NCEP Modeling Synergy Meeting Highlights: March 31, 2008


The meeting was led by Michael Brennan and attended by Mary Hart, Geoff DiMego, Ed Danaher, Jim Hoke, Joey Carr, Chris Magee, Jeff Ator, Mark Iredell, Zoltan Toth, John Ward, Keith Brill, and Hendrik Tolman.  Stephen Jascourt of UCAR/COMET, Pete Browning of Central Region HQ, David Bright and Steven Weiss of SPC, and Steven Silberberg of AWC attended by remote teleconference.


1. NCO


John Ward gave a rundown of the following upcoming implementations:


  • The NAM and Great Lakes Wave model upgrades were implemented with the 12 UTC cycle today.
  • John reported that he had heard back from most of the NCEP service centers concerning the evaluation of the new GFS implementation. Steve Weiss from SPC said they would provide a limited evaluation as they are entering the peak severe weather season at this time of year. John also stated they would like as much participation as possible from the regions in this evaluation.
  • The upcoming GFS implementation is running in parallel and output will become available on ncosrv on April 1. The parallel GFS will not run at the same time as the operational model due to computing resource limitations. The parallel GFS output will be available about 5.5 hours after that of the operational model.
  • A CCS upgrade will be completed in about three weeks, with production switching from Mist to Dew on April 2, then back to Mist.
  • A planning meeting with the service centers scheduled for April will discuss the next CCS upgrade along with a potential moratorium on changes for the 3rd and 4th quarters of FY09.
  • Operational implementation of the Alaska RTMA is scheduled for April 15.
  • Stephen Jascourt asked about an email he had seen discussing a compiler bug which causes wrong answers to certain matrix-multiply calculations in a developmental version of CRTM radiation code used in the GSI.  When compiled with a new compiler from IBM, the answers are correct.  At this point in time, it is not clear if this bug has any adverse effect on operational codes. The new compiler will be the default compiler on the new Power6 based computers coming in the future.


A detailed list of all upcoming upgrades and implementations can be found here.




2a. Global Climate and Weather Modeling Branch


Mark Iredell reported on the following:

·        Testing is underway for a GFS upgrade this summer that will incorporate several new data types.

·        By late summer/early fall, a GFS product implementation is planned to clean up scripts related to the post processing and add new products to the GFS output. If there are any new product requests, please make them soon as these changes will require 75-day change notice.


2b. Mesoscale Modeling Branch


Geoff DiMego reported on the following:


  • The NAM implementation became operational today, March 31.
  • The next NAM bundle, currently being tested, includes partial cycling which uses T–12 atmospheric fields from the GFS rather than from the NDAS. Testing results since December have shown consistently improved performance with this change, mostly with the representation of larger scale features. Also being tested is the incorporation of TAMDAR data, which has shown a small positive impact.
  • Re-instatement of fire weather IMET support runs is still being explored. NCEP had proposed nesting down to 2.67-km grid spacing in the vicinity of a fire, however, there is no easy way to deliver this output to WFOs via AWIPS. Given this, EMC would like feedback from the regions on the relative importance of these fire weather runs compared to work on the next NAM bundle; dedicating resources to the reinstatement of the fire weather runs would likely delay the delivery of the next NAM bundle into FY09.
  • Code has been completed for several small changes to the hi-res window runs, including hourly output, the immediate generation of GEMPAK output files as each output time becomes available instead of waiting until the end of the run to begin the GEMPAK processing, and the computation of updraft helicity. These changes will not require a full parallel evaluation.
  • Hi-res window output will be available in AWIPS OB9.
  • Feedback is requested from the NCEP service centers on how they are using the CONUS RTMA and centrally-produced SmartInit downscaled grids.
  • There is a need for evaluation feedback from the regions on the SmartInit grids (the “Downscaled NWP Grids” OSIP project).


2c. Global Ensemble Prediction System


Zoltan Toth reported on the following:


  • Dataflow from FNMOC has been established via NCO, allowing receipt of FNMOC ensemble output. Currently, the FNMOC ensemble has 16 members, and is run only at the 00 UTC cycle. The FNMOC ensemble will be evaluated for potential inclusion in NAEFS in two years.
  • ECMWF ensemble data is also under evaluation for potential inclusion in the NAEFS in two years.
  • In the 4th quarter, an implementation is planned to increase the grid resolution of the GEFS, to implement hind-casting, stochastic perturbations, and increase the length of the run from 16 to 30 days for one cycle per day.
  • Future GFS upgrades will no longer be simultaneously implemented in the GEFS. Instead, these changes will be incorporated into the GEFS during the next scheduled ensemble system upgrade.


2d. Short Range Ensemble Forecast System


Geoff DiMego reported on the following:


  • A SREF implementation will begin running in parallel in the 4th quarter of FY08 or the 1st quarter of FY09 to upgrade the WRF members to the latest available version and increase the resolution of the RSM and WRF members to 32 km (to match the Eta members).  Also, the components making up the SREF will be adjusted to include more WRF members. The new SREF will be composed of:

            6 Eta members

            5 RSM members

            10 WRF members (5 NMM, 5 ARW)


2e. Marine Modeling and Analysis Branch


Hendrik Tolman reported on the following:

  • The Great Lakes Wave Model upgrade forced with NAM winds was implemented today, March 31
  • The Great Lakes Wave Model forced with NDFD winds will be tested in the parallel slot vacated by the above implementation.  There is no timeframe to make this version of the model operational, but output will be made available in AWIPS OB9.
  • A wave model ensemble implementation, scheduled for May, will increase the number of members from 10 to 20 and increase the run length from 5 to 10 days. Work is underway with FNMOC to unify grids and model versions for the ensemble. Other changes include running off of bias-corrected ensemble winds and using initial conditions for each member cycled from that same member in the previous run. This results in the generation of perturbations for the entire wave field, not just those waves generated by the winds in the current run. The branch will work with users who want a more robust dataflow to evaluation this upgrade.
  • The next upgrade to HYCOM is awaiting a parallel slot to test new data processing that significantly improves the representation of the Gulf Stream and other features.




Steve Weiss of SPC asked where GFS parallel output will be located. NCO replied that it would be in the usual “gfsp” slot on ncosrv.  Steven Silberberg AWC asked about the format of the DBNET alerts for the parallel data, and this will be clarified.


Jim Hoke of HPC reported that QPF performance scores of the NAM and GFS for some thresholds for FY08 were the lowest in the last 6 or 7 years. However, the scores improved markedly in February to some of the highest scores ever seen for that month.


Pete Browning of Central Region stated that the Minneapolis WFO is evaluating the SmartInit downscaled grids and will provide feedback to EMC.


4. The next Synergy Meeting will be held Monday, April 28, 2008, at 12:00 pm EDT in Room 209 at EMC, with remote conference capability.