Backgound

Understanding weather modeling: An In-Depth guide to weather models and their accuracy

Weather modelisation is a crucial process in predicting the state of the atmosphere using advanced computer software. By solving the mathematical equations of thermodynamics, weather models forecast atmospheric conditions based on data from weather stations, balloons, radars, and satellites.
June 28, 2024
3-5 minutes

A weather model is a computer software that solves the mathematical equations of thermodynamics to predict the state of the atmosphere at a given moment. These equations have been known for over 200 years and are used to express the physical behavior of the fluid that is the atmosphere, as well as the energy exchange processes between the atmosphere, the earth, and the oceans. Weather models rely on parameters measured at each moment by various instruments (weather stations and balloons, radars, satellites, etc.).

Each model has a different level of precision that depends on the computing power allocated to solving the equations. For calculations to be perfectly accurate, it would be necessary to provide information at the given moment for each molecule of the concerned fluids with absolute precision, without any measurement error. This is impossible!

To simplify, we divide the atmosphere into small cubes whose size depends on the grid (in degrees) and the number of vertical levels. For each grid point, the calculation is performed to track all atmospheric parameters (temperature, rain, humidity, wind, etc.). The larger the grid of a model, the faster the calculation, but the less precise the result.

Developing a classic weather model is long and costly. As a result, there are about fifty models currently in existence, with 10-15 standing out with usable results. The most well-known are:

GFS (Global Forecast System, NOAA)

  • One of the most used in the world; 4 times a day
  • Resolution: 20km
  • Advantages: forecasts up to 15 days, everything is free
  • Disadvantage: generalist model, mediocre local quality for weather parameters. It gives the trend.

CEP by ECMWF (European Centre for Medium-Range Weather Forecasts)

  • Resolution: 9km
  • Advantage: good quality, large scale, 10-15 days for trends
  • Limitation: the finest resolution is paid, difficulties with ultra-localized phenomena

ARPEGE (Action de Recherche Petite Echelle Grande Echelle, Météo-France)

  • Resolution: 7.5km over France, 37km in remote regions
  • Advantage: covers the entire planet
  • Limitation: difficulties in forecasting ultra-localized phenomena

AROME (Application of Research to Operations at Mesoscale, Météo-France)

  • Resolution: 1.3km
  • Advantage: precision
  • Limitation: only for a very restricted territory (France and partly neighboring countries) and short-term (36 to 48 hours)

ICON (Icosahedral Nonhydrostatic, DWD)

  • Resolution: 13km, 7km, and 2km (Germany)
  • Advantage: good forecasts for 72 hours over Europe and particularly Germany, every hour
  • Limitation: difficulties in forecasting ultra-localized phenomena

Ensemble models:Unlike the deterministic ones mentioned above, instead of sticking to one scenario, weather forecasters use ensemble models; initial conditions are slightly disturbed by adding a bit of error; this allows for a range of scenarios and therefore the probabilities of certain phenomena occurring.