Numerical models based on the fundamental equations of hydrodynamics may be used to forecast the weather, based on observed initial conditions. These models are run on high-speed computers at the National Center for Environmental Prediction. Numerical forecasting models provide guidance in forecasting many weather parameters, including pressure fields, humidity, clouds, winds, temperatures, precipitation, and sea heights. Models forecast at scales including global, synoptic, and regional. They fall into two general categories: those based on statistical data concerning frequency of occurrence (i.e., of thunderstorms, tropical storms, tracks and intensities) and those based on hydrodynamic equations of the behavior of the atmosphere.
These "dynamic" models rely on input data that can be from observations (ship observations, station observations, bouy data, radiosonde data, etc.) or from the outputs of other models. Different dynamic models represent the atmosphere at different levels of resolution in terms of horizontal scale or "grid size" (i.e., 9 kilometer, 27 kilometer) and the number of vertical layers of the atmosphere across which the calculations run. Models also differ in their temporal range (i.e., short-range,. medium-range, and long-range).
Models differ in that some take terrain (elevation) into account. Some models also accept data on sea surface temperatures. The run of a dynamic model begins with a process called "initialization" in which the parameters must be set for the initial conditions in each cell and for the conditions at the boundary of the model's areal extent. Thus, for example, the output from a global model can be used to specify the conditions of the atmosphere at the boundary for a regional model.
The initial conditions for the cells within the region would be initialized on the basis of observational data. After initialization, the model runs the hydrodynamic equations iteratively across intervals (e.g., 6 hours, 12 hours, 36 hours, etc.). All models have biases and tendencies--things they tend to forecast well, things they do not, things they tend to under-forecast, things they tend to over-forecast, etc.
Models that are widely used in the weather forecasting community include:
· The Aviation Run (AVN) of the National Center for Environmental Prediction's Medium-Range Forecasting Model
· Naval Ocenaographic Global Atmospheric System (NOGAPS)
· The Global Model run by the United Kingdom Meteorology Office (UKMET)
· The model run by the Geophysical Fluid Dynamics Laboratory (GFDL)
· The Climatology and Persistence (CLIPER) statistical model of hurricane track forecasting run by the National Hurricane Center,
A variety of web pages describe models and their tendencies in detail.