Meteorological forecasting models are effectively used in a variety of commercial and military applications. To improve accuracy, forecasts can be made using fusion of more than one meteorological model. Data fusion can reduce the noise in the prediction and overcome problems caused by intermittent data losses. Adaptive data fusion enables a forecast system to adjust to changes in model skill and weather regimes. Weights associated with each input forecast module represent the relative skill of the inputs.

Numerous schemes have been developed for the calculation of these weights. These techniques generally use a history of forecasts and observations in the weight determination process. The relative skill of each fusion technique depends heavily on the data being integrated. This paper evaluates the skill and computational complexity of several of these weight determination techniques in a meteorological application.