| 摘要: | 劇烈天氣降水為台灣春夏交際時常見的天氣事件，並且容易對經濟造成損失，甚至對於大眾的生活以及安全造成危害；因此當劇烈天氣發生時該如何有效應對以及警戒是一項重要的議題。本研究使用中央氣象局用於預報作業上之對流監測平台系統(System for Convection Analysis and Nowcasting, SCAN)當中以對流胞辨識與追蹤(Storm Cell Identification and Tracking, SCIT)方式所得出的對流胞資料，進行2015年～2018年總共4年來5月到8月之統計分析。由於台灣主要為東北──西南走向之地形，天氣系統在北部以及南部皆有不同的表現，因此北部採用五分山雷達之觀測資料，南部則以七股雷達資料將台灣地區分為北部以及南部進行分析探討。藉由統計上所得到的資訊定量決定對流胞預報過程中之預警範圍。進一步利用2019年5～8月資料進行驗證。
在對流胞特性分析上，北部地區之對流胞多為自陸地往海洋或在洋面上移動，南部地區主要是自海面往陸地上移動；由於挑選月分為春夏交際之時，因此大多數對流胞還是以西南向東北進行移動。生命週期方面絕大多數對流胞落在1小時內，其次為1～2小時；在位置分布上，北部區域分佈於海面上之對流胞數量相較於南部地區海域對流胞數量來得多，而分布密度上不論南北區域依然是陸地區域之對流胞密度高於海面區域。
而對流胞預警方面，本研究採用類似於颱風路徑潛勢預報(PTA)的方式將對流胞統計過去誤差後取前70%之最大誤差，並將對流胞依照4種速度進行區分，不同速度對應至不同誤差範圍，挑取個案進行預報測試；結果顯示整體命中率約為60～70%。由於SCIT演算法本身的限制，對流胞初筆資料並不會有移動方向以及速度以進行外延預報，因此後續加入了其他系統所提供之中尺度環流場進行測試，初步測試結果為使用MAPLE運動場在大多數個案可以達到近似於原本SCIT資料直線外延之效果，甚至在部分個案表現出之效果比原先結果來得更好，因此推測可以補足SCIT上不足之處，以增加劇烈天氣警戒能力。;Severe weather system accompanies with heavy rainfall is a common event which often occurs in summer time over Taiwan area. It affects people security, life safety, and the economy. Therefore, how to predict such kind of events and prevent the disasters is an major issue for the operational unit. In this study, the Storm Cell Identification and Tracking (SCIT) algorithm in the System for Convection Analysis and Nowcasting (SCAN), which is applied in the forecast center of Central Weather Bureau (CWB), is used to survey the occurrence of convective cells, identify their locations and track the movements for the nowcasting. To examine the convective cells statistically, a set of historical data between March and August from 2015 to 2018 are selected for this study. In addition, the RCWF (Wu-Fan San) radar data is collected to represent northern Taiwan area, and the RCCG (Chi-Gu) radar data is selected to represent southern Taiwan area.
Results of the analysis show that, most of convective cells have about 1-hour life time. In addition, the distribution of cells shows more convection events over the ocean area in northern Taiwan compared with southern Taiwan, but the cell density over land area is higher than ocean area in both location.
To improve the nowcasting quantitatively, the tracking error of the convective cells are estimated in statistics. Based on the moving speed of the cells, the Potential Track Area algorithm is applied in 4 categories to define the affected radius of convective cells in the period of 0-1h nowcasting. By examining the performance of storm tracking in 2019, result shows that the hit rate is about 60~70%. When further providing additional information of environment flow, the performance of nowcasting can be further improved in some cases. |
