簡單統計方法於台灣地區颱風降水預測之研究（一）預測方法與台北颱風降水之預測校驗

A Study of Typhoon Rainfall Statistics Forecast over Taiwan Area Part 1: Methods and the Forecast Evaluations over Taipei

葉天降(Tien-Chiang Yeh) ； 吳石吉(Shih-Chi Wu) ； 謝信良(Shinn-Liang Shieh)

《大氣科學》 27卷4期 (1999/12) Pp. 395-412

摘要

颱風降水預報是台灣地區天氣預報與災害防範作業中，最重要的工作項目之一，王等（1986）所謂平均法之預測結果，是中央氣象局在颱風降水預報作業裡，主要的客觀參考資料。本研究以1961至1996共36年更完整之颱風降水資料，校驗平均法在台北颱風降水預測之準確度，同時探討引用持續性於颱風降水預測之可行性。研究的結果顯示，平均法對颱風降水之預測有基本之參考價值，以西行颱風為側，時降水量預測值與觀測值間之相關係數（Rxy）為0.42，均方根誤差（Rms）為3.75mm/h，而對24小時以內之累積降水量預測結果之Rxy值都在0.49以上，平均時降水量Rms值都在3.01mm/h以下，如6/12/24小時累積降水量預測結果之Rxy值為0.56/0.59/0.42 Rms值為2.67/2.36/2.15。但平均法預測則對颱風較大降水個案明顯低估，以持續法預測颱風降水，主要只能改善較短時距之預測，對大於六小時以上時距之預測結果，基本上都較以平均法之預測結果為差。我們進一步綜合平均法與持續法，發展所謂差異持續法與比差法。由台北颱風降水預測結果之校驗顯示，此兩方法能改善對颱風較大降水之預測能力，提高Rxy值也不致使Rms值明顯增加。以西行颱風為例，對3/6/12/18小時累積降水預測結果之Rxy值已可分別提高為0.61/0.61/0.62/0.54，而對3/6/12/18小時累積降水達到25/75/150/200mm較大降水之公正預兆指數值（Equitable Threat Score）則可達0.22/0.23/0.25/0.22，對18小時以內之累積降水量預測已有作業參考價值。對北行颱風，由於多數資料個案位於較南方，在台北地區降水較小，使各預測方法之預測準確度降低，但仍顯示差異持續法與比差法能改進平均法對颱風較大降水低估的情形。

關鍵字

颱風 ； 降水預測

並列摘要

The forecast of typhoon rainfalls is one of the most important and challenging task of the weather forecast and disaster mitigation during typhoon season in Taiwan area. The method from Wang et al. (1986), which is based on climatology average and is called 'Climatology Average' method, has been adapted to the operational usage of typhoon rainfall forecast in Central Weather Bureau. In this study, we evaluate the performance of the Climatology Average by using a more complete data set including 36-year rainfall observations over Taipei from 1961 to 1996. Examining of forecast by using the idea of persistence is also conducted. The results show that the Climatology Average can provide a basic guidance for the Taipei typhoon rainfall forecast. For the hourly rainfalls forecast of westward moving typhoons, the correlation coefficient between the forecasts and the observations (Rxy) is 0.42, and the root mean square error of the forecasts (Rms) is 3.75 mm/h. For the forecast of 3-/6-/12-/18-/24-hour accumulated rainfall, the Rxy is greater than 0.49, and the Rms is smaller than 3.01 mm/h . However, all the heavier rainfall cases are under estimated by the Climatology Average. The forecast of assuming the rainfall remained persistent, Persistence Forecast, can improve the forecast only for very short time ranges. The Persistence Forecast is generally less accurate than the Climatology Average when the time ranges are greater than 6 hours.Combining the Climatology Average and the idea of persistence, we developed two simple statistics rainfall forecast methods of DEviation PeRsistence (DEPR) and DEviation STandardized (DEST). The evaluations show that the performances of both methods of DEPR and DEST are similar and both can improve the forecast of the heavier rainfall. The Rxy value for the forecast of the 3-/6-/12-/18-hour accumulated rainfall of westward moving typhoons can be increased to 0.61/0.61/0.61/0.54. And, the equitable threat score with threshold of 25/75/150/200 nun accumulated rainfall in 3-/6-/12-/18-hour can be increased to 0.22/0 .23/0.25/0.22. For northward moving typhoons, the majority cases are located farther south to Taipei. The rainfalls in Taipei are smaller and the performance of the rainfall forecast is slightly degraded. However, similar improvements are found of forecasts from DEPR and DEST.

並列關鍵字

Typhoon ； Rainfall forecast