Weather Forecasting with Ensemble Methods.

The article focuses on developments related to the practice of numerical weather prediction in the U.S. Until the early 1990s, atmospheric scientists viewed weather forecasting as an intrinsically deterministic endeavor: For a given set of best input data, one best weather prediction is generated. Armed with sophisticated computing resources, weather centers ran carefully designed numerical weather prediction models to produce deterministic forecasts of future atmospheric states. Weather prediction has been transformed through the implementation of ensemble forecasts. An ensemble forecast comprises multiple runs of numerical weather prediction models, which differ in the initial conditions and/or the numerical representation of the atmosphere, thereby addressing the two major sources of forecast uncertainty. The ability of ensemble systems, in concert with statistical post-processing, to improve deterministic forecasts--in that the ensemble mean forecast outperforms the individual ensemble members--and to produce probabilistic and uncertainty information to the benefit of weather-sensitive public, commercial, and humanitarian sectors has been convincingly established.