A generalised Model Output Statistics (MOS) system for Australian applications has been developed. It is general as to location and predictand, and while a fairly comprehensive set of potential predictors is currently available, the grid-point array nature of the development data set makes it simple to add further potential predictors. The system is applied to develop seasonal equations for the prediction of maximum and minimum temperature at Australian capital cities, and it is shown that these equations are substantially more skilful than the first operational MOS equations, which were developed on an unstratified data set.
Mills et al. (1986) studied this question.