The Weather Research and Forecasting (WRF) Model and Data Assimilation System

The WRF-system effort in NESL/MMM supports an extensive WRF community that includes academic, operational, and commercial interests. WRF model components now include: the Advanced Research WRF (ARW) solver, the Non-hydrostatic Mesoscale Model (NMM) solver, WRF-Chem, the Advanced Hurricane research WRF (AHW), WRF-Fire, the Nested Regional Climate Model (NRCM), and the Global ARW. The WRF-Var data assimilation system includes: 3-dimensional variational (3DVAR), 4-dimensional variational (4DVAR), the WRF Data Assimilation Research Testbed (DART) ensemble assimilation system, and nudging methods.

WRF support activities include overall management of the WRF system, incorporating community enhancements and contributions into the model, regular WRF system updates, maintenance of a help service, and national and international workshops and tutorials each year. A key part of the support effort includes ensuring that WRF code improvements/upgrades are readily accessible to other configurations for specific applications; NESL/MMM staff maintains a single-source code repository so the latest versions of all components are available in the most recent release. In addition, the WRF team undertakes WRF-based research and development, provides support for field projects using real-time WRF, and conducts fundamental research on algorithms and techniques for the application of ensemble data assimilation. Substantial experimental and theoretical research is conducted on understanding physical and chemical processes to support development of improved methods for incorporating such processes in WRF. Close collaboration with universities and other NCAR scientists is a feature of this research and applications work. The WRF-Chem model is used for field program analysis, and to investigate regional-scale air quality and interactions between meteorology and chemistry. NESL scientists contribute to development of the WRF-Chem code by implementing new processes related to gas-phase and aerosol chemistry and adding diagnostic variables for the current model. The scientists evaluate and develop the model with several types of simulations, ranging from field program analysis of megacities to chemistry-climate studies. Close collaboration with other NCAR scientists and universities, particularly University of Miami, University of Washington, Penn State University, Yonsei University in Korea, and University of Oklahoma, is a feature of this research and applications work.

In FY 2011 NESL/MMM staff conducted two tutorials in Boulder, one in January–February and another in July.  The first was attended by about 65 people and the second by about 60–80 people.  MMM also organized and hosted the 12th WRF Users’ Workshop in Boulder June 20–24. The first day presented a continuation of the model fundamentals lecture series that was begun last year, with the topic being PBL and land surface modeling.  The body of the workshop featured WRF topic areas, while the final day offered instructional talks.  Attendance was about 230.

MMM released WRF Version 3.3 in April 2011, offering many new features and improvements.  They reflect contributions from throughout the WRF community and speak to WRF’s position as a community model.

  • Cumulus/convection options: Tiedtke cumulus scheme, new SAS (Simplified Arakawa-Schubert) scheme from the GFS, Zhang-Macfarlane scheme from CESM, University of Washington shallow convection scheme
  • Microphysics: SUNY–Stonybrook (5-class) microphysics scheme, updated Thompson scheme, updated Milbrandt scheme, modified Morrison scheme
  • Radiation: Goddard shortwave and longwave radiation physics schemes, improved RRTMG longwave scheme PBL and surface layer options: University of Washington TKE scheme, Total-Energy Mass Flux (TEMF) PBL
  • HWRF improvements
  • Numerous new WRF-Chem capabilities
  • WRFDA upgrades: New RRTOV radiative transfer model, redesigned 4DVAR, new control variable option

MMM once again ran the ARW at 3-km resolution to support the NOAA Storm Prediction Center’s Spring Forecast Experiment (SFE).  As in years past, the forecasts were provided to the experiment participants and to NWS forecasters.  This year, for the first time, the simulations were initialized with analyses produced by an ensemble data assimilation system using NCAR’s Data Assimilation Research Testbed (DART).  This employed a 50-member ensemble to generate the 15-km analyses used, for the experiment period of April 27–June 12.  The effort was reported on at the 12th WRF Users’ Workshop, with results showing competitive performance compared to previous seasons’ forecasts using more traditional approaches.

AMPS 12-hour forecast
Figure: AMPS 12-hour forecast for area around NSF R/V Nathaniel B. Palmer.  Surface winds (barbs; full barb = 10 kt), sea level pressure (contoured), and 3-hourly precipitation (mm, shaded) shown.  Location of ship marked “NBP.”  Edge of Ross Ice Shelf to south, with Marie Byrd Land coastline to east (rightward).

MMM continued to run the Antarctic Mesoscale Prediction (AMPS) System, a real-time WRF system that provides guidance for forecasters of the United States Antarctic Program (USAP).  AMPS employed WRF to complete its support of the 2010-2011 USAP field season.  Developed this past season was a capability to produce WRF plots to track the course of NSF’s R/V Nathaniel B. Palmer as it cruised the Ross Sea and Southern Ocean.  The figure below presents an example of one of the ship-following WRF products.  This shows the Palmer’s location (labeled “NBP”) north (above) of the Ross Ice Shelf ice edge and west (leftward) of the Marie Byrd Land Coast, with surface winds, sea level pressure, and 3-hourly precipitation plotted.

MMM also again ran WRF in real-time over the Atlantic Basin for the 2011 hurricane season.  The model was set up with a 3-grid configuration with 36-, 12-, and 4-km spacing.  This real-time system was run for Hurricane Irene, the hurricane which caused extensive flooding in the Northeast.  The figure below shows a sample plot of forecast central pressure and wind speed for Tropical Storm Philippe, for a five-day forecast initialized at 1200 UTC 26 Sept 20111.

Tropical Storm Philippe
Figure: Forecast central pressure (mb) and wind speed (ms-1) for Tropical Storm Philippe from real-time WRF hurricane run initialized at 1200 UTC 26 Sept 2011.