Postprocessing
The use of statistical techniques to transform numerical weather prediction (NWP) output (possibly ensemble output) into a prediction of a meteorological variable or variables, often with the aim of improving meteorological guidance and decision-making. The transformation may incorporate data from sources outside the NWP (e.g., observations or climatology) and can take many forms. For example, it may include a first-moment correction (i.e., bias correction) or a correction of higher moments (e.g., ensemble resampling or dispersion correction), or it may produce the forecast of a derived variable not explicitly part of the NWP output (e.g., fog or maximum wind over some period of time). Postprocessing can be a multistep process and involve a variety of statistical techniques (e.g., regression analysis or discriminant analysis). Most postprocessing techniques fall into the categories of model output statistics (MOS) and perfect prog (PP).
Term edited 2 March 2020.
