A set of different forecasts all valid at the same forecast time(s). The differences between the forecasts can provide information on the probability distribution of the predicted variables. The forecasts in the ensemble may have different initial conditions, boundary conditions, parameter settings, or may even be from entirely independent NWP models.