A method or technique of developing objective forecasting aids. Suitable statistical relationships are found between a predictand and one or more observed variables that can be forecast by one or more numerical (dynamic) prediction models. The relationships can be determined by linear or nonlinear regression, multiple discriminant analysis, or other statistical methods. In practice, the relationships are applied to the appropriate output of numerical prediction model(s) to yield forecasts of the predictand. In essence, the output of the model(s) is considered perfect, hence the name. The difference between model output statistics (MOS) and perfect prognostic is that in MOS the predictand is related to the actual model output, while in perfect prog, the predictand is related to observations or representations of them at (nearly) concurrent times.