NCEP Modeling Synergy Meeting Highlights: December 31, 2007


This meeting was led by Michael Brennan and attended by Keith Brill, Mary Hart, Geoff DiMego, Ed Danaher, Bill Bua, Hendrik Tolman, and Ken Campana. Stephen Jascourt of UCAR/COMET and Rusty Billingsley of Southern Region HQ attended by remote teleconference.


1. NCO


No representative of NCO was present. A detailed list of all upcoming upgrades and implementations can be found here.




2a. Global Climate and Weather Modeling Branch


Ken Campana reported on the following upcoming global branch implementations:


A new GFS upgrade will be discussed by the EMC implementation board on 10 January with code to be delivered to NCO shortly thereafter. This upgrade will involve changes to the GSI, a new quality control scheme, upgrades to the shortwave and longwave radiation physics and the enthalpy replacing virtual temperature as the thermodynamic state variable (due to the model extending beyond the mesosphere). Implementation is tentatively scheduled for 1 April. 


This upgrade may impact downstream users of the native GFS “sigma” data, and these users may need to modify their software again. These users will be notified and an updated GFS "users’ manual" will be made available.


Another minor GFS upgrade is planned for the fall, largely to allow for the assimilation of new data types.


2b. Mesoscale Modeling Branch


Geoff DiMego reported on several issues related to MMB:


The change to the enthalpy formulation in the GFS will impact the model output and has the potential to impact downstream users of the GFS data, including the NAM.


A NAM bundle submitted in late November includes:


  • An increase in the domain size by 18% (in the north, west, and east directions), this will have downstream impacts for users of the native NAM grid output (e.g., SREF, Hi-Res Window, RUC, etc.)
  • An upgrade in the i,j,k WRF-NMM code that will allow the model to run more efficiently and offset most of the increase in run time due to the expanded domain
  • Introduction of gravity-wave drag and mountain blocking
  • An upgrade to the NOAH LSM to match the NCAR version
  • An upgrade of the GSI to the August 2007 version, however the strong balance constraint will not be included
  • A slight modification of the positive definite advection scheme


Geoff has created a pair of tables highlighting differences between the GFS and NAM and their respective data assimilation systems that are available online at:

Geoff also reported that although in many cases the performance of the NAM improves when the model is re-run with GFS initial conditions, is it not feasible to initialize the NAM with those data operationally due to time constraints (as too much satellite data arrives too late for the NAM to wait for the GFS analysis) and the domain configurations.


However, it is feasible to use an older global model state in the NDAS cycle.  This partial cycling is being tested in which atmospheric fields at the beginning of the assimilation period 12 hours before the model cycle time will use a 6-hour global model forecast from the GDAS valid 12 hours prior to the cycle time instead of a 3-h WRF-NMM forecast valid 12 hours prior to the cycle time from the last NDAS cycle.  For land surface states, full cycling will continue to be used, due to problems initializing the NAM LSM from GFS data.  The assimilation will still follow the usual NDAS procedure of running the NAM-GSI to assimilate data at 12, 9, 6, 3, and 0 hours before the cycle time and running the NAM for the intervening 3-hour segments to supply the first guess.  This will test whether using the global model atmospheric conditions for the first guess in the t-12 hour analysis yields the benefits seen from running the NAM off GFS initial conditions while providing 12-hours of spin-up for mesoscale features to adjust to the NAM resolution and terrain. 


The implementation of NAM downscaled grids has been delayed 30 days by NCO. These grids are created using SmartInit and will have the following resolution: CONUS: 5 km, Alaska: 6 km, Hawaii: 2.5 km


2c. Global Ensemble Prediction System




2d. Short Range Ensemble Forecast System


A SREF upgrade scheduled for the 2nd Quarter of FY08 will include additional bias corrected output, but this output will not be delivered to the field immediately.


2e. Marine Modeling and Analysis Branch


Bandwidth issues are causing the delay of up to 20 minutes in the delivery of the high resolution wave model output.


The Great Lakes wave model code has been delivered to NCO with a scheduled implementation for February 2008. This upgrade will use new software and provide additional output (e.g., wind wave and swell) but will not include major changes to the model itself.  A parallel test running the model using NDFD winds is also planned.


Work is underway to sync the NCEP wave ensemble forecast system with that of FNMOC.


Issues with the location of the Gulf Stream continue with HYCOM. This problem may be improved with changes to the data assimilation, but this will require a shorter model time step, possibly delaying the receipt of the model output. However, this might not be a problem since RTOFS Atlantic data need to be available later for coupling with HWRF.


Work to couple HWRF with HYCOM instead of the POM will continue, but this will not be operational for the start of the 2008 hurricane season. It is hoped that the HWRF coupled with HYCOM will be available in parallel for the hurricane season.




WFO Reno


Stephen Jascourt passed along a question from the Reno WFO about a disparity in elevation between GFS forecast soundings from the 40-km AWIPS 212 grid and soundings from BUFR files. The entire sounding appears shifted to lower elevation in the BUFR sounding than on the 212 grid. Stephen will send an example of this problem to Ken Campana and Mark Iredell for further investigation.


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