History Ensemble forecasting

as proposed edward lorenz in 1963, impossible long-range forecasts—those made more 2 weeks in advance—to predict state of atmosphere degree of skill owing chaotic nature of fluid dynamics equations involved. furthermore, existing observation networks have limited spatial , temporal resolution (for example, on large bodies of water such pacific ocean), introduces uncertainty true initial state of atmosphere. while set of equations, known liouville equations, exists determine initial uncertainty in model initialization, equations complex run in real-time, use of supercomputers. these uncertainties limit forecast model accuracy 6 days future.

edward epstein recognized in 1969 atmosphere not described single forecast run due inherent uncertainty, , proposed stochastic dynamic model produced means , variances state of atmosphere. although these monte carlo simulations showed skill, in 1974 cecil leith revealed produced adequate forecasts when ensemble probability distribution representative sample of probability distribution in atmosphere. not until 1992 ensemble forecasts began being prepared european centre medium-range weather forecasts (ecmwf) , national centers environmental prediction (ncep).