Ensemble forecasting is a numerical prediction method that generates a representative sample 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 set of observations, or measurements. 

The multiple simulations are conducted to account for the three main sources of uncertainty in weather forecast models:

1. Errors introduced by chaos or sensitive dependence on the initial conditions;
2. Errors introduced by imperfections in the model, such as the finite grid spacings. 
3. Errors introduced by incomplete  initial conditions (which are varied slightly for the different simulations).
   

Ideally, the verified weather pattern should fall within past ensemble spreads, and the amount of spread should be related to the probability of certain weather events occurring.