Advances in the Numerical Method and Technology of Multi scale Numerical Prediction

With the advances of computer and computational technique, multi scale multi task numerical model is being developed for fine prediction in a global scope. The difference between a general circulation model and a mesoscale model becomes unclear except for the forcing arrangement, and a global model tends to be used for severe storm prediction and study of mesoscale convective systems. Development of a multi scale model can be stated in improving the universality of model dynamics, discretion scheme, computational algorithm and physics, and the key issue is improving the multi scale properties of each parts. It calls for no approximation to the dynamics of the real atmosphere and detailed arrangement of the topography. Numerical schemes and the computational algorithm should be flexible, accurate and stable, so that can be used to various grid spacing systems. Considering the spherical mesh, grid structure and dynamical feature of discretion, the quasi uniform mesh with fine numerical dispersion feature is preferred in the recent development of numerical models. Reduced grid, cubed grid, icosahydron grid, Yin Yang grid and Voronoi grid are discussed. In addition to detailed model physics, non hydrostatic model with global and regional configuration and unified frame is the fashion of multi scale model. The vertical coordinate and terrain arrangement are important for proper description of dynamical and thermo dynamical processes. Vertical z coordinate and “shaved cell” topography have boosted the unified model, especially in high resolution case. Besides, numerical method should describe the intrinsic nature of dynamical processes, and provides sub grid information during the temporal integration. Conservation of the main model properties is known as the key factor for long term integration. High performance numerical scheme favors the multi scale application of a model and is of great attraction. The pioneer researches over the world are summarized, which provides reference for the development of the multi scale model system, GRAPES, in China.