The Application of Model Output Statistics to Precipitation Prediction in Australia

Abstract

The Model output Statistics (MOS) technique has been used to produce forecasts of both the probability of precipitation and the rain amount for seven major Australian cities in subtropical and middle latitudes. Single station equations were generated using data from the current objective analysis, together with some surface observations for the same time, and a 24 h prognosis based on that analysis, to predict the rainfall in the 24 h beyond the prognosis validity time. In order to increase the usefulness and acceptability of the MOS predictions, transformations were applied that reduced the biases of the final forecasts throughout the forecast ranges. The skid with which large rainfall totals were predicted was particularly enhanced in this manner the MOS forecasts showed much greater skill in the prediction of large totals than was achieved by either the operational or persistence forecasts, while predicting small totals with comparable proficiency. The MOS probability forecasts were better able to predict rainfall occurrence than were the quantitative MOS forecasts, and additionally were superior in this regard to both subjective forecasts produced operationally and predictions based on persistence. The overall skill of the quantitative precipitation forecasts was further enhanced by using the probability estimates to provide a categorical prediction of rain occurrence such that a rain amount was only forecast if the predicted probability of precipitation exceeded 50%. Routine issuance of the MOS guidance to the operational forecasters commenced in January 1984.