Considerable experience exists with applications of nonhydrostatic models to simulations of meso scale processes. However, numerical weather prediction (NWP) deals with motions on a much wider range of temporal and spatial scales. Difficulties that may not be significant on the small scales, may become important in NWP applications. Based on these considerations, a new approach has been applied in developing the NCEP Nonhydrostatic Meso Model (NMM) within the WRF effort. Namely, instead of extending the cloud models to synoptic scales, the hydrostatic approximation is relaxed in a hydrostatic model formulation. In this way the validity of the model dynamics is extended to nonhydrostatic motions, the number of prognostic equations remains the same as in the hydrostatic model, and at the same time the favorable features of the hydrostatic formulation are preserved. This approach does not involve any additional approximation.

In the model, “isotropic” horizontal finite differencing is employed that conserves a variety of basic and derived dynamical and quadratic quantities. Among these, the conservation of energy and enstrophy improves the accuracy of nonlinear dynamics of the model. In the vertical, the hybrid pressure-sigma coordinate has been chosen as the primary option. The forward-backward scheme is used for horizontally propagating fast waves, and an implicit scheme is used for vertically propagating sound waves. The Adams-Bashforth scheme is applied for non-split horizontal advection of the basic dynamical variables and for the Coriolis force. In real data runs the nonhydrostatic dynamics does not require extra computational boundary conditions at the top.

Since recently, the NMM has been run operationally at NCEP. The efficiency of the computational algorithm of the model significantly exceeds the efficiency of algorithms of most state-of-the-art nonhydrostatic models. In high resolution NWP applications, the model has been highly competitive with mature hydrostatic NWP models and with other nonhydrostatic models. Examples illustrating the performance of the model are presented.