- 期刊
應用同步衛星資料估算臺灣地區對流降雨
Applying Satellite Data to Estimate Convective Rainfall
摘要
伴隨降雨的潛熱釋放是熱帶地區大氣的主要能量來源之一，因此透過雨量的估算，可以瞭解熱帶地區能量或水汽的變化及傳遞情形。而由於海洋佔了熱帶地區的絕大部分，所以降雨的估算在氣象測站極度缺乏且分布極度不均的熱帶地區，應用具有空間連續且時間解析度非常好的同步衛星資料是極為重要的方法之一。一九九一年六月底的連續幾天豪雨，造成臺灣地區的火水成災，損失之巨為歷年所罕見。本文即利用該時段時同步衛星GMS-4的紅外線與可見光影像資料，配合臺灣地區地面雨量觀測網測得的雨量資料，研究找出衛星影像資料與降雨間的相關性，並建立估算降雨模式以及估算模式的最佳區域面積，同時亦研究驗證對流降雨特性在臺灣地區是否會因地區不同而有所不同。此外，以可見光影像濾除卷雲所造成的誤判結果亦是本文探求的目的之一。本研究結果顯示，若以五個統計值（平均亮度溫度，標準偏差，最低亮度溫度，溫度小於230K的雲面積，以及溫度小於210K的雲面積）為參數的迴歸模式，則其與平均降雨率的相關係數為51.63%，而經以可見光濾除卷雲後的資料計算，則其相關係數升為56.43%。在以未經可見光濾除卷雲的資料所建立的模式之驗證中，發現估算中雨的準確率為59.5%，估算大雨的準確率為53.6%，而若經可見光濾除卷雲後的估算準確率在中雨與大雨的類別上分別增為59.7%與56%，小雨部分由於資料數少而難以以其結果為準確率之代表。研究中亦發現最佳估算面積大小應為10*10像元的面積，而由模式的驗證中亦可知臺灣地區的對流降雨特性會因地區的不同而互異。
並列摘要
Precipitation is a very important factor in energy transport in tropical area. The change and transport of energy or water vapor can be understood through the estimation of precipitation. Because water surface is the most part of tropical area, the weather station is very scarce in this region. Therefore, one of the best methods used to measure the rainfall is applying the geostationary satellite data. It is not only good in special coverage, but also good in temporal continuity.The heavy rainfall that occurred in June 21-24, 1991 made Taiwan area in flood and caused a very huge damage. This study applied the infrared and visible data observed in the same time period by the geostationary satellite, GMS-4, and combined with surface rain gauge data to find out the optimal model of rainfall estimation. The optimal area size of rainfall estimate for satellite data application and the precipitation characteristic of different region for convective rainfall were diagnosed in this study. Using visible data to filter out cirrus influence was also discussed.The research results showed that the correlation coefficient between the average rainfall rate and the regression parameters is 51.63% if the mean brightness temperature standard deviation, minimum brightness temperature, cloud area where the temperature is colder than 230K, and cloud area where the temperature is colder than 210K are used as regression parameters. The correlation coefficient will increases to 56.43% if data is applied to filter out cirrus cloud. The percentage of correct estimation is 59.5% for middle rainfall rate, and is 56.6% for heavy rainfall rate. If visible data are applied to filter out cirrus cloud, the percentage of correct estimation will increase to 59.7% and 56%, respectively. Because the samples are scarce, it hasn't the verification for light precipitation in this research. The results also showed that the optimal area size for rainfall estimation is 10×10 pixel size and there existed local characteristics in convective precipitation in Taiwan area.