Numerical Weather Prediction

Numerical Weather Prediction Computer simulation of atmospheric processes starting from the actual state is called numerical weather prediction. Physical laws can describe the basics of such a simulation, development of weather. Mathematical formulation of these laws leads to a system of mainly non-linear partial differential equations so that analytical solution is not possible.

These equations can be solved approximately, using numerical methods.
For that purpose atmospheric parameters are defined in a finite number of numerical points, horizontally and vertically, creating grid boxes representing area of interest.
Starting from the initial state, calculation of temporal values of relevant parameters is proceeding by solving equation system in every time step for every grid box.
Value of time step depends of horizontal resolution as well as numerical methods used in calculation. Generally, there are two types of numerical models used for weather prediction, global and regional.
The global models cover the whole globe while the regional models are for a limited area. Atmospheric processes are extremely variable, especially in lower layers; therefore their exact presentation needs a very high resolution of numerical grid.

Due to limited computer recourses and reasonable operational time for model running, the resolution of numerical models depends on weather phenomena that are a desirable feature. For example, distribution of high and low pressure in the atmosphere has a characteristic dimension of about 1000km while thunderstorms are of importance in areas of several km. One way to overcome this problem is by using models for limited area with high resolution and the global model with coarser resolution for preparing the lateral boundary conditions.