NLT3D: a non-local three-dimensional turbulence parameterization scheme for next generation numerical weather prediction models

点击翻译按钮获取中文摘要

Numerical weather prediction models have to split the atmospheric processes into two classes: processes on spatial scales larger than the grid size can be simulated explicitly by the basic equations (conservation laws), while subgrid scale processes have to be represented by physical parameterization schemes. With increasing computer power many national forecast centers now aim at horizontal grid sizes in the kilometer and subkilometer range to be operationally available in the next years. Since most parameterization schemes used in forecast models have been developed for horizontal grid sizes of several kilometers and more, their formulation and selection of processes included will then become questionable. This holds especially for the turbulence parameterization representing subgrid scale dynamics and transport mainly in the planetary boundary layer (PBL). Since here the exchange of energy, momentum, moisture and other trace gases with the earth's surface takes place, a correct simulation of the PBL processes is decisive for a good forecast quality. Since the most energetic large eddies in the boundary layer cannot be represented by classical K approaches simulating local mixing by small eddies, non-local (but still one-dimensional) turbulence schemes have been developed to account for smaller vertical grid sizes. On the other hand, small grid sizes in the horizontal direction now also claim the representation of horizontal turbulent exchange between the grid boxes, i.e. the simulation of three-dimensional (3D) turbulence. To overcome these problems, we plan to develop a non-local parameterization scheme for 3D turbulence by extending an existing non-local one-dimensional turbulence parameterization by a horizontal turbulent exchange. As a model environment for development and testing of our scheme we will use the WRF model from UCAR/NCAR, USA. However, the algorithms in the scheme will be formulated in a way allowing for implementation in other high resolution forecasting models as well.