In the last three years, NCAR and the Army ATEC Program have jointly developed a real-time multi-grid (with grid sizes of 0.5 - 45 km) rapid-cycling four-dimensional data assimilation and forecast (RTFDDA) system. The system was deployed and has been running operationally at five Army testing ranges. The high-resolution, accurate, real-time analyses and forecasts from the RTFDDA system have become a dependable tool in the daily operational meteorological support for tests at these Army test ranges. Considerable effort was spent at continuously improving the data assimilation and the model physics of the system. This paper will report on some improvements to the model PBL simulation, through augmentation of the surface flux computation and fine-tuning the urban-soil property parameters.

Verification of the surface analyses and forecasts of the RTDDDA systems indicated that the systems tend to systematically underestimate the daytime surface winds for most seasons. This occurred at all the Army ranges, which are located in very different geographic and climatological areas. The FDDA analysis could corrected the wind bias, but the correction does not survive into a longer forecast. Recently, a few other investigators recognized a similar problem in their case simulations and/or operational MM5 forecasts with the broadly-used Hong and Pan (1996) (commonly referred to as MRF) PBL scheme (e.g. Zhang and Zheng, 2003). It was speculated that the model might be underestimating the downward momentum flux in the free-convection PBL during daytimes. Our recent analysis with RTDDDA model output reveals different reason. We found that the MRF model appears to significantly overestimate the convection-induced turbulent mixing under free convection, through the so-called convective velocity (w*). The excessive w* results in an overestimate of the friction velocity (u*) and imposes excessive friction from the ground surface. It is this excessive surface friction that severely reduces surface momentum and leads to the excesssively weak surface winds. To solve the problem, a more sophisticated physically-based Beljaars' (1994) surface flux formulation was implemented into the MRF PBL scheme. In the new formulation, the w* is computed directly from the model surface sensible and latent heat fluxes, and the PBL height. A new approach to define the PBL height with the local Richardson-Number is also implemented and compared with the non-local (bulk) Richardson-Number method. It is found that the PBL height obtained with the new method greatly mitigates the known overestimation of PBL heights in the original MRF scheme. In addition, the Zilitinkevich (2001) surface heat flux formulation is introduced. Simulation experiments with the RTFDDA model configuration indicate that these modifications not only effectively reduce the daytime weak wind bias, but also improve the surface thermal and moisture forecasts. Case studies show that the revised MRF scheme performs comparable to the much more expensive ( ~25% of the total model run-time) TKE-based ETA Mellor-Yamada PBL scheme.

The revised MRF PBL scheme was employed in the RTFDDA operation for the Joint Urban 2003 (JUT) Field Experiment in Oklahoma City during late June and July. Five nested grids with a fine grid size of 500 m were designed to properly resolve the OKC areas. To appropriately simulate the urban effect, urban parameters of the NOAH land-surface model were adjusted to better reflect the large heat capacity, small water storage capacity and other properties of the urban surface. Both case studies and statistics of JUT operational runs show reasonable simulation of the urban heat island and its effect on modification of the PBL structure and local circulations in conjunction with the revised MRF PBL and the urban parameter refinements.