NUMERICAL WEATHER PREDICTION MODELS

So what are numerical weather models and how do they work?

Numerical models attempt to simulate the state of the atmosphere at various times in the future. Models need to ingest data before they begin running. Where do they get their data from? Well, every day at synoptic times 12Z and 00Z worldwide, radiosonde balloons are released that collect weather observations up to 100,000 feet (30 kilometers). Radiosondes are units that measure various atmospheric parameters, such as air temperature, humidity, and pressure, which are transmitted to a fixed receiver on the ground. Models are initialized using observed data from the radiosondes, weather satellites, and surface weather observations.

The forecasts are computed using mathematical equations for the physics and fluid dynamics of the atmosphere. These equations are nonlinear and are impossible to solve exactly. As a result, numerical methods obtain approximate solutions.

Different models use different solution methods. Some global models (GFS,  ECMWF) use spectral methods for the horizontal dimensions and finite difference methods for the vertical dimension, while regional models and other global models usually use finite-difference methods in all three dimensions. Regional models (NAM) also can use finer grids to explicitly resolve smaller-scale meteorological phenomena, since they do not have to solve equations for the whole globe. The irregularly-spaced observations are processed by data assimilation and objective analysis methods, which perform quality control and obtain values at locations practical by the model’s mathematical algorithms (usually an evenly-spaced grid).

Primitive equations are initialized from the analysis data and rates of change are determined. The rates of change predict the state of the atmosphere a short time into the future. The equations are then applied to this new atmospheric state to find new rates of change, and these new rates of change predict the atmosphere at yet a further time into the future. This time stepping procedure is continually repeated until the solution reaches the desired forecast time. Forecast time maxima for global models may range from 10-16 days (GFS) whereas regional/mesoscale models forecast ranges from a few hours to 4 days (NAM).

Is there more than just the operational model run? Read about ensembles ——>