- 期刊
簡單統計方法于臺灣地區臺風降水預測之研究（二）預測結果隨區域之分布
A Study of Typhoon Rainfall Statistics Forecast over Taiwan Area Part Ⅱ:Spatial Distribution of the Forecasts
摘要
葉等(1999)曾以1961至1996共36年之降水資料，詳細校驗平均法（見王等1986）、持續法、差異持續法與比差法在台北颱風降水預測之情形。本文主要在延續葉等(1999)之研究，以相同時期之資料與相同之方法，校驗這些統計方法在台灣其它地區颱風期間之時、6小時累積與12小時累積降水預測情形。 從颱風降水資料分析顯示，台灣地區颱風降水和地形有相當密切的關係，各地較大降水發生時機是當颱風中心移到颱風之氣旋式環流可吹向地形，當地處於迎風面時發生。從平均法對降水量的預測也顯示，迎風面地區之颱風降水預測準確度較背風面地區之預測準確度為高，顯示地形強迫作用有助於在颱風來襲時產生系統性的降水。 對颱風降水進一步校驗之結果也顯示，平均而言北行颱風之降水較西行颱風降水少，且以上述統計方法做北行颱風降水預測之準確度較對西行颱風降水預測之準確度為低。對所檢驗各地區中，不論西或北行颱風皆以西南部高雄與台南地區之颱風降水預測誤差最大，北行颱風則也使西北部台中與新竹地區之颱風降水預測誤差相對擴大。所應用之降水平均場若依颱風路徑分為西行與北行兩類，並沒有增進颱風時降水量與6小時累積降水量預測之準確度，但對12小時累積降水之預測，則顯示應用依路徑分類之平均場，較應用不分類之平均場，所做預測為準確。此外，應用平均法進行颱風降水預測，因採用平均值，因此對較大降水都有低估的現象，引用差異持續法或比差法對許多地區能改進其部分缺失，因而提高颱風降水預測之準確度，其中又以對6小時累積降水預測之改善較為明顯。
並列摘要
Based on a 36-year typhoon rainfall data set from 1961 to 1996,Yeh et al.(1999)evaluated the Taipei typhoon rainfall forecasts of some statistical methods. The methods used include the Climatology Average, the Persistence, the Deviation Persistence method and the Deviation Standardized method. This study uses the same data set and applies the same approach to evaluate the typhoon rainfall forecast over the other areas of Taiwan. The observations show that the typhoon rainfalls over Taiwan area are significantly affected by the Taiwan orography. Heavy rainfalls usually observed over the up-slope area of the typhoon cyclonic circulation impending the mountains. The higher correlation between the forecasts and the observed rainfalls are also in the up-slope area. The orographic forcing has an effect to induce the systematic rainfalls. The forecast evaluations show that the averaged rainfall is smaller, and the correlation between the forecasts and the observed rainfalls is also smaller, for the northward moving typhoons than for the westward moving typhoons. Among the areas examined, the largest forecasting errors are found at Kaohsiung and Tainan over Southwestern Taiwan. For northward moving typhoons, the forecasting errors at Hsing-chu and Taichung over Northwestern Taiwan are also large. The evaluations also show that separating the rainfall mean field into westward and northward categories based on the typhoon moving direction does not improve the hourly rainfall forecasts. Similar conclusions are applicable to the 6-hour accumulated rainfall forecast. However, for the 12-hour accumulated rainfall forecast, the forecasting errors are smaller when two categories of mean fields were applied. Under-estimations of the heavy rainfalls are also found from the forecasts of the Climatology Average method. Both the Deviation Persistence method and the Deviation Standardized method are able to reduce such kind of bias and to improve the forecast, especially for the 6-hour accumulation rainfall forecast.