On FDDA / model physics for improved 4-km mesoscale model simulations for air-quality applications

A previous study showed that use of analysis nudging four-dimensional data assimilation (FDDA) and improved physics in the MM5 produced the best overall performance on a 12-km domain simulation. However, further reduction of the grid length to 4 km in the simulation had detrimental effects on the meteorological solutions. The primary cause was traced to the explicit representation of convection accompanying a cold frontal system. Because no convective parameterization was used on the 4-km grid, the convective updrafts were forced on coarser-than-realistic scales, and the rainfall and the atmospheric response to the convection were too strong. The evaporative cooling and the associated downdrafts were too vigorous causing widespread disruption of the low-level winds and spurious advection of the simulated tracer.

A series of MM5 and WRF experiments was designed to address this general problem involving 4-km model precipitation and grid-point storms. The MM5 part of the work has been completed, and results associated with 4-km grid model sensitivities to grid explicit microphysics, convective parameterization, planetary boundary layer (PBL) turbulence physics, enhanced horizontal diffusion, and use of analysis nudging versus observation nudging FDDA have been written up and submitted for publication (Deng and Stauffer 2005). Some of the conclusions from this study include: 1) Enhanced parameterized vertical mixing in the Turbulent Kinetic Energy (TKE) turbulence scheme has shown marked improvements in the simulated fields, 2) Use of convective parameterization on the 4-km domain improved the precipitation and low-level wind results, 3) Use of the MRF PBL scheme showed larger model errors within the PBL and a clear tendency to predict much deeper PBL heights than the TKE scheme, 4) Combining observation nudging FDDA with a convective parameterization on the 4-km domain produced the best overall simulations, 5) Finer horizontal resolution does not always produce better simulations, especially in convectively unstable environments, and new convective parameterizations suitable for 4-km resolution are needed, 6) Although use of current convective parameterizations at 4-km resolution may violate their underlying assumptions related to the size of the convective element relative to the grid size, the grid-point storm problem was greatly reduced by applying convective parameterization to the 4-km grid.

In the current study, some of the aforementioned conclusions regarding 4-km grid sensitivities to model physics and FDDA will be re-tested with the WRF Eulerian Mass (Advanced Research) dynamic core, and compared to those using the MM5. A converter code has been developed with NCAR help to convert pressure-level MM5 initial condition fields, lateral boundary conditions, and analyses for FDDA, to the WRF infrastructure. The WRF experimentation is currently underway and some WRF and MM5 results will be compared at the symposium.