Improvement and testing of WRF physics options for application to Rapid Refresh and High Resolution Rapid Refresh

Since the last ARAM meeting in January 2010, there has been intensive testing and evaluation of the Rapid Refresh (RR), which uses the Weather Research and Forecasting (WRF) Advanced Research WRF (ARW) model, in preparation for operational implementation. The implementation schedule calls for the RR to replace the Rapid Update Cycle (RUC) later this year as the primary National Centers for Environmental Prediction (NCEP) model source for forecasts of high-impact aviation weather over the CONUS. The RR domain will cover all of North America (including Alaska) and adjacent ocean areas with 13km horizontal grid spacing. The RR will update with an hourly assimilation cycle, same as the RUC.

In addition, work toward improving explicit prediction of convection and mesoscale convective systems by the High Resolution Rapid Refresh (HRRR) has greatly accelerated since additional computing resources became available in the fall of 2010. HRRR will cover CONUS with a storm-resolving 3-km horizontal grid, and use the WRF-ARW model to make forecasts every hour out to 15 hours. Initial and lateral-boundary conditions for the HRRR are now provided by the RR, so there is a close coupling between the initial conditions for the HRRR and the data assimilation used in the RR. For both the RR and HRRR, a two-pronged and strongly coupled effort has concentrated on 1) the data assimilation for the RR and on 2) the physics in the WRF-ARW model. This paper will focus on the latter.

For the RR, we are using the Goddard short-wave and the Rapid Radiative Transfer Model (RRTM) long-wave radiation schemes, the RUC Land-Surface model (RUC-LSM) for prediction of soil temperature and moisture, and snow cover to give surface fluxes, the Mellor-Yamada-Janjic (MYJ) surface and boundary layer parameterization for turbulent vertical fluxes by sub-grid eddies, the Grell "G3" scheme for deep and shallow convection, and the Thompson microphysics for grid-scale cloud and precipitation. Motivated by the need for a better performance over the Arctic, and toward overcoming problems with snow ablation that result in systematic temperature biases, as well as a need to improve our wind forecasts at low levels for wind-energy forecasting applications, we have - added sea ice into the RUC LSM, including accumulation and ablation of snow on the ice; - modified treatment of snow over land areas to improve evolution of the snow cover, including melting; - extensively modified, tested and evaluated the Mellor-Yamada-Nakanishi-Niino (MYNN) surface and boundary-layer schemes as possible replacements for the MYJ.

For the HRRR, we are using the same physics as for the RR, except without parameterization for convection. Here our main physics-related activities have been directed toward overcoming deficiencies in HRRR forecasts for CoSPA noted during the 2010 convection season. These include insufficient propagation and sometimes insufficient longevity of mesoscale convective systems (MCSs), and delayed initiation and poor coverage of weakly sheared convective storms over the southeast US during summer. In addition to the work toward modification of the MYNN boundary and surface-layer schemes noted above, which stands to benefit low-level wind forecasting by the HRRR for wind-energy applications, we have determined that use of the 6th order diffusion was partly responsible for the deficiencies of the HRRR in predicting southeast US convection, and are exploring the role of the microphysics as a source of poor performance on prediction of MCSs.

This research is partially in response to requirements and funding by the Federal Aviation Administration (FAA). The views expressed are those of the authors and do not necessarily represent the official policy or position of the FAA.