Ensemble forecasting is a numerical prediction method that generates multiple representative calculations of the possible future states of the weather.
Multiple numerical predictions (model runs) are conducted using slightly different initial conditions that are all plausible given the past and current sets of measurements and observations.
The multiple simulations (see pictures) are conducted to account for the three main sources of uncertainty in weather forecast models:
- Errors introduced by incomplete initial conditions (which vary slightly for the different simulations).
- Errors introduced by chaos or sensitive dependence on the initial conditions.
- Errors introduced by imperfections in the model, such as the finite grid spacings.
Ideally, the verified weather pattern should fall within the range of ensembles, and the frequency of the distribution should be related to the probability of certain weather events occurring.
Ensemble forecasts are mainly applied to large-scale weather conditions for the forecast range of 3 to 14 days.
Applications to local forecasting in high resolution are limited due to the very high computing effort required, and to the large influence of large-scale patterns on regional developments over a time of more than 5-7 days ahead.