嘉南地區1997年梅雨期降水之特性與作業模式預測校驗

The Characteristics and Model Forecast Evaluations of Rainfalls in Chiayi-Tainan Area during 1997 Meiyu Period

葉天降(Tien-Chiang Yeh)

《大氣科學》 26卷1期 (1998/03) Pp. 41-61

摘要

豪大雨預報是現今氣象預報工作中最重要而富挑戰性的工作，為期能對豪大雨預報和其延伸之災害防護上有進一步突破，國內氣象界組成嘉南地區定量降水整合研究，本文即在相同目標下，探討中央氣象局有限區域預報模式之降水預測，應用在梅雨期嘉南地區0－24小時雨量預報之適用性。本文主要採用1997年5月及6月間之觀測與模式預測之每12小時累積降水量，模式預測結果包括60公里解析之粗格組（CA）及20公里解析之細格組（FA）在預測時間0－12小時、12－24小時及24－36小時等三段12小時累積降水量。校驗時是將預測格點資料內插到嘉南地區之嘉義、台南和阿里山等三觀測站之位置和其觀測值比較。研究之結果顯示中央氣象局有限區域預報模式中FA之降水量預測較CA之預測結果好，而綜合應用三時段預測結果（相當於時延系集預測）要較僅使用單一時段之預測結果要好。在CA和FA結果相近時，綜合使用CA和FA之結果也較單一使用CA或FA要好。我們也發現1997年梅雨期時在嘉南地區之降水量有相當明顯的日夜變化，模式之預測也能顯示這種趨向。以使用FA三時段降水之平均值預測嘉南地區是否將發生降水之ETS（Equitable Threat Score）值在日夜間為0.28在日間為0.36。對日間是否發生大於20mm降水預測（有限個案）則以應用FA 12－24小時及24－36小時二時段之平均值為最佳，其ETS值為0.26。

關鍵字

降水 ； 預報校驗

並列摘要

Heavy rainfall forecast is the most important and challenging task of weather forecast in Taiwan area. Heavy rainfalls in the area often end up with floods and disasters. A term works focus on the improvement of the quantitative precipitation forecast in Chiayi-Tainan area has been carried out since 1995. This study, with the same purpose of improving rainfall forecasts, investigates the opportunity of applying the outputs of the Limited Area Forecasting Model, Central Weather Bureau (CWB) to do rainfall forecast in Chiayi-Tainan area.Current version of the Limited Area Forecasting Model, CWB includes a coarse resolution mesh (CA) of grid distances 60 km and a fine resolution mesh (FA) of grid distances 20 km. Evaluations of the model predicting rainfalls are in the data period of May and June (so call meiyu period) 1997. The model predicting 12h accumulated rainfalls are interpolated into the location and compared with the observations at three CWB stations of Chiayi, Tainan, and Alishan in Chiayi-Tainan area.Our results show that the rainfall intensity in Chiayi-Tainan area have very clear diurnal variations during 1997 meiyu period. The model predicting rainfalls are able to catch such a phenomenon. More detail evaluations show that the performance of FA rainfall predictions is better than that of CA. By applying time lag averaging 'ensemble forecast' can generally improve the forecast. The equitable threat score (ETS) of predicting rainfall events by using the mean of three 12h forecasts is about 0.28 (night time) and 0.36 (daytime). And, the ETS of predicting greater than 20mm rainfall events for limited number of daytime cases is about 0.26.

並列關鍵字

Rainfall ； Forecast evaluation