Numerical Weather Prediction models at FMI

Sami Niemelä

Numerical weather prediction (NWP) models provide the basis for modern weather forecasting by simulating the evolution of the atmospheric state. The core of the atmospheric model is the set of hydrodynamic equations, such as the conservation laws for mass, motion, heat, water (in solid, liquid and vapour forms) and any other gaseous (chemical) species. Numerical methods are utilised to solve the above set of conservation laws represented by coupled partial differential equations averaged over a finite grid. In addition, the effect from the flow structures and processes that are not represented by grid-scale motions must be taken into account via physical parametrization methods. The fundamental idea behind the physical parametrization procedure is to estimate the sub-grid-scale terms by using the averaged prognostic variables. Parametrization processes estimate the net effect of the sub-grid scale flow structures such as the fluxes momentum, heat and moisture due to small turbulent eddies and larger convective secondary circulations. Moreover, treatment of radiation, cloud microphysics and interactions between the atmosphere and the surface, are also needed in atmospheric models.

NWP-application is an initial value problem where the model integration is critically dependent on the initial conditions. In addition to the precise atmospheric model, a successful forecast requires that both the upper air and the surface initial conditions provided to the forecast model are accurate and consistent with available observations. Furthermore, the essence of any NWP-system is to produce both the initial state and the forecast as fast as possible, however, still maintaining the reliability and accuracy of the code. Therefore, the scaling and code efficiency in massively parallel HPC-systems are the critical success factors for NWP-applications.
This presentation will give a short overview on the NWP-models used operationally at FMI. Furthermore, the special attention is given to the new mesoscale NWP-model Harmonie.