Operational Model Output Statistics (MOS) forecasts of maximum and minimum temperature and precipitation quantity and probability of occurrence for seven Australian capital cities have been prepared daily since 1 October 1983. The levels of accuracy of the first twelve months of these forecasts are described, and an attempt is made to partition the sources of error in these forecasts into errors in day to day prognoses, deficiencies in the development data set, and errors inherent in the MOS approach. By means of a one-month test it is demonstrated that the MOS forecasts are sensitive to changes in the error characteristics of the numerical weather prediction model used to provide the predictors. It is concluded that the factors which would have the greatest positive impact on the accuracy of MOS forecasts are Improved accuracy of the day to day prognoses, and stratification of the development data by season to provide separate equations In each season for each predicted element.
Tapp et al. (Sun,) studied this question.