92nd American Meteorological Society Annual Meeting (January 22-26, 2012)

Tuesday, 24 January 2012: 4:15 PM
Data Assimilation Improvements for the 2nd Version of the Rapid Refresh
Room 340 and 341 (New Orleans Convention Center )
Stephen S. Weygandt, NOAA/ESRL/GSD, Boulder, CO; and M. Hu, S. G. Benjamin, H. Lin, C. R. Alexander, D. C. Dowell, and P. Hofmann

The initial version of the Rapid Refresh (RR) mesoscale analysis and prediction system has been frozen and is planned for operational implementation at NCEP in the fall of 2011. Like the Rapid Update Cycle (RUC), which it will replace, the RR is a high frequency (1-hour) cycling system that provides short-term weather forecasts, that are used for aviation, severe weather, and general weather forecasting guidance. The RR utilizes the Gridpoint Statistical Interpolation (GSI) for the analysis component and the Advanced Research WRF (ARW) for the model component.

Extensive testing and refinement of specific features within the GSI and the coupled GSI/ARW system for Rapid refresh application has yielded an initial RR system that represents a significant improvement in forecast skill compared to the RUC system. Areas of refinement for the RR have included use of a partial cycle with information from the GFS added in twice per day, fine tuning of GFS background error covariance information, incorporation, inclusion of a RUC-like cloud analysis package within the GSI, and modification to the surface assimilation to better account for differences in elevation between the model surface and the METAR observations.

Work on a further set of data assimilation improvements for the 2nd version of the RR (slated for operational implementation late in 2012) has been ongoing since summer of 2011. Several enhancements are being examined, including: 1) use of mesonet and radar radial velocity data with improved data quality control, 2) nudging of surface temperature and moisture fields based lowest atmosphere analysis increments, 3) refinement of the background error covariance matrix (create directly from RR 1h forecasts instead of GFS 6h forecasts), 4) modification of the GSI to better use surface observations and better fit radiosonde and aircraft observations (use of finer radiosonde vertical levels, reduced observation error variance and vertical background correlation length scales), 5) enhancements to the cloud/hydrometeor analysis and radar-DFI application, and 6) improvements to the satellite assimilation method for the RR.

At the conference, we will describe the work in these areas and its impact on the RR forecast skill. Time permitting, we will also summarize preliminary work toward designing a prototype regional hybrid EnKF system for the RR.

Supplementary URL: