Research And Application Of MOS-Based Objective Precipitation Forecast Model

With the method of Out Put Statistics, Cluster Analysis and Possibility Function of Precipitation, a new approach is carried out on the improvement of objective quantitative precipitation forecast by using numerical forecast productions of NMC T213 model from June to August in 2003~2006. Precipitation observation is prepro-cessed using relative humidity before forecast model is made, and then by picking up predicttors of T213 numerical forecast productions and clustering, forecast model of cluster objective precipitation is developed. Three schemes are designed to make the operational prediction experiment on the precipitation of nine stations of Jiang-huai area in July of 2007.Details of schemes are show as below: nation non-cluster precipitation forecast model(scheme1), nation cluster precipitation forecast model(scheme2), Jiang-huai area cluster precipitation forecast model(scheme3).According to the three meteorological categories after clustering, objective precipitation forecast model is used and the forecasting results of three types are compared with observations on several stations of Jiang-huai area in July,2007.It shows that the coming 36 hours precipitation can be mainly revealed by each scheme, but there is still difference on quantity of precipitation between forecast and observation.Sche-me3 has the best accuracy of light rain forecast among three schemes.By verifying and analyzing the forecast result of each scheme, it's indicated that the 36h light rain forecast result of scheme3 is better than other schemes in every aspect ,such as TS score, no-hitting rate, fault-hitting rate, forecast bias. Especially absent forecasts are reduced. The 36h middle rain forecast result of scheme2 is the best through TS score; meanwhile scheme3 still has the best result of 60h middle rain forecast. Analyzing results of three schemes, the method of clustering associated with model output statistics(MOS) presented in the article is a new test to fully utilize a large mount of the numerical forecast production. It's a effective way to improve forecast level and technique.