透過您的圖書館登入
IP:18.116.42.208

摘要


本文針對台灣大學、中央大學、台灣師範大學、中國文化大學、中央氣象局及民用航空局在2000年及2001年梅雨季的12-24小時、24-36小時MM5系集預報的降水產品,利用全台343個自動雨量站之降雨觀測資料,以ETS值及BIAS值進行校驗,藉以評估此系集降水預報系統的預報能力,並試著找出最適合台灣地區的系集降水預報組合方式。 個別來看,六個系集成員中以GR的表現最佳,可見以Grell積雲參數化法與Resiner I 微物理過程的組合在六組中最適用於台灣地區梅雨季之降水預報。 整體而言,以系集平均方式產生的降水預報得分約在第二、三名之間,並非最佳之預報。比較2001年各權重方式與系集平均的結果發現權重方式確實能提升降水預報能力,在B時段以系集weight1的預報成績最佳,C時段則次於系集weight2。系集weight1係利用各模式在A時段的預報成績以決定B、C時段各模式的權重,可見模式在A時段的表現與B時段的相關性較高,不過一旦模擬時間增長至C時段,系集weight1的預報能力便開始下降。系集weight2係利用2000年各模式的預報成績以決定2001年的權重,其結果在B、C時段皆有不錯的預報得分,可見依過去各物理組合表現予以權重的方式亦能提升降水預報之能力。

並列摘要


This study presents precipitation verification, in the Taiwan area, for a real-time MM5 ensemble forecast system during the 2000 and 2001 Mei-yu seasons. The forecast system consists of 6 MM5 models with different physics settings that are separately run by 4 universities(NTU,NCU,NTNU,CCU)and 2government institutes(CWB,CAA)in Taiwan. It is found that the GR model outperformed the other 5 members. Therefore, it is strongly recommended that MM5 simulations in the Taiwan area during Mei-yu seasons should consider using Grell cumulus scheme and Resiner I microphysics process(the GR model).Generally speaking, the score of precipitation forecast from ensemble mean is the second or the third highest within all the members. In addition to the ensemble mean, we also use 2 weighting methods to produce ensemble precipitation forecasts. In ensemble weight1,the weighting of each ensemble member for computing ensember rainfall at12-24 and 24-36 hours is determined according to the performance of its 0-12 h rainfall forecast.Ensembleweight2 utilizes the ETS score of rainfall forecast of each member in 2000 to obtain its weighting for ensemble rainfall forecast in 2001.The result shows that the 2 ensemble methods help increase the accuracy of ensemble precipitation forecasts. They both in genernal outperformed the ensemble mean and each single member.

延伸閱讀


國際替代計量