整合氣象雷達與即時降雨資料於颱風降雨推估之研究

Abstract

本研究利用中央氣象局QPESUMS (Quantitative Precipitation Estimation Segregation Using Multiple Sensor) 系統之雷達估計雨量進行雨量特性分析，整合地面雨量站觀測，再透過數值與統計分析方法，求得具有代表性的定量降雨估計 (QPE) 資訊，可以提供各類依賴雨量進行預測災害警戒訂定之參考。雷達觀測可以反應空氣中水氣含量，雖無法精準估計出空間中水汽量的多寡，但是可以做為地面雨量估計的參考。而地面雨量站的降雨觀測資料是反映該地實際降雨的情況。但是自動雨量觀測站之設置成本頗高。因此，無法在地面大量架設自動觀測雨量站。所以，對於地面降雨的分布情形就得依賴各種內插方式來求得。本研究嘗試將高解析度的雷達降雨估計資料與地面雨量觀測資料共同整合，以求得出較精確的地面降雨估計。從案例研究成果發現，Kriging 方法在差值校正上的表現較Barnes為佳，因此可運用於未來災害即時預警之降雨指標，以改善目前因降雨估計誤差所導致災害預警的不確定性。

This research adopts the radar detection, updated every 10 minutes, from the QPESUMS (Quantitative Precipitation Estimation-Segregation Using Multiple Sensor) system provided by the Central Weather Bureau for analyzing rainfall characteristics and integrating this data with the surface rain-gauge observations. A representative Quantitative Precipitation Estimation (QPE) will be obtained through numerical and systematic analysis, which can be applied as a reference for precipitation-based disaster forecasts and warning-issue standards. The major advantage of radar detection is the precision of spatial analysis. The 1.3 km X 1.3 km grid of the radar data is much finer than rain gauge observation. Furthermore, radar can estimate moisture content in the air; it can be used as a reference for surface rainfall estimates. Rain gauges only report the pinpoint data at very limited location. Therefore, estimations of surface rainfall distribution are obtained by a series of interpolation methods. From extensive case study results, the Kriging is more effective than Barnes. Therefore, it can be used in the future for precipitation standards during instant disaster warnings to improve the uncertainty of the errors in the current rainfall estimation.