84th AMS Annual Meeting

Tuesday, 13 January 2004: 3:45 PM
Results of WRF development activities at NCEP
Room 605/606
Geoff DiMego, NOAA/NWS/NCEP/EMC, Camp Springs, MD; and T. Black, H. Chuang, J. Derber, S. G. Gopalakrishnan, Z. Janjic, D. Keyser, M. Pyle, and W. Wu
NCEP's Environmental Modeling Center has been active in developing many aspects of the end-to-end WRF Modeling system, including observation preparation, 3-D VAR analysis, initial and lateral boundary conditions, dynamic cores, physics modules, post-processor, verification, ensembles and operational implementation. Status and highlights of these activities will be presented to demonstrate the breadth and commitment of NCEP to the overall WRF effort. For context, the NWS goals for implementing WRF capability in NCEP Operations begin in 2004 with a WRF ensemble in the High Resolution Window run - currently 8 km Nonhydrostatic Mesoscale Model (NMM), followed in 2005 by WRF components being implemented in the North American Mesoscale Early run - currently 12 km Eta, then in 2006 the Rapid Refresh WRF - currently 20 km RUC and Hurricane - currently moving nest of 18 km GFDL and, finally, in 2007 in the Short Range Ensemble Forecast system.

In early 2003, NCEP began running a test end-to-end WRF system based on the Eulerian mass-core coupled with a set of WRF physics chosen to match the physics running in NCEP's NMM. Periodic forecast runs were made out to 48 hours that mimicked NCEP's Operational High Resolution Window runs which are made daily with NMM over four large area fixed domains and two small fixed domains nested within the Eta which supplies initial conditions and hourly one-way lateral boundary conditions. The large domain runs are made at 8 km for eastern CONUS (18z), 8 km for central CONUS (12z), 8 km for western CONUS (06z) and 10 km for Alaska (00z). The small domain runs are made at 8 km for Hawaii (00z & 12z) and at 8 km for Puerto Rico (06z & 18z). Identical forecast output have been generated from the WRF runs and subjected to objective verification versus observations and precipitation analyses for comparison with the corresponding NCEP runs of the NMM.

In August 2003, in association with the WRF Implementation Test Plan, NCEP will begin making real-time runs for two nests each day. The purpose of these runs is to evaluate the two control configurations being planned for the WRF Ensemble. One control is identical to the NMM already running in the HiResWindow runs while the other will be the Eulerian mass-core with a physics configuration chosen by NCAR to be most suitable for that dynamics core. Both runs have to fit into a similar computer resource / time slot. With its higher-order numerics, the Eulerian mass-core is more costly to run so its grid-spacing was increased to 10 km to get its runtime to approximate that of the NMM. The higher-order accuracy is believed to ameliorate any degradation that coarser resolution might cause when compared to the 8 km NMM. Keep in mind that these runs are not intended as a "bake-off" but rather to establish the two WRF cores and their respective physics as sufficiently accurate and of comparable quality for use in the WRF ensemble (which will have a total of 6 members).

The purpose of making runs in real-time is to provide users an opportunity to participate in the subjective evaluation of the model guidance provided by the two control runs. Websites will be setup with side-by-side displays and evaluation forms will be provided to gather forecaster feedback opinions. The Alaskan nest and central CONUS will be run from 1 September through 30 October while western CONUS and eastern CONUS will be run from 1 November through 15 December. Objective statistics, a summary of the subjective evaluations and sample cases will be presented.

Supplementary URL: