| 摘要: | 由於台灣降雨既豐沛又極端，準確的定量降雨估計顯得格外重要。未來台灣將由S及C波段雙偏極化雷達組成雷達降雨估計網，而雙偏極化雷達參數會因其波長差異而有不同的特性，這特性將影響其應用於降雨估計的結果。此研究目的為分析不同波段雷達其降雨公式的表現，以利於未來各波段雷達之定量降雨估計網的整合。
本研究使用設置於國立中央大學(NCU)測站長達6年以上之二維雨滴譜儀資料，依月份分類降雨類型(春雨、梅雨、午後熱對流、颱風、冷鋒面)後經T-matrix散射模擬雙偏極化雷達參數，並回歸三波段雷達各降雨類型之雷達參數-降雨關係式(係數)，分別為R-Z、R-KDP、R-(Z,ZDR)及R-(KDP,ZDR)。研究個案選自2014年3月至2015年8月8日蘇迪勒颱風前，由NCU C-Pol與RCWF S波段雷達共同觀測之個案，共9個案累計20小時。將雷達觀測參數進行資料品管後套用至各降雨公式，並使用地面96個雨量站資料評估雷達降雨估計之表現。
在未考慮雷達觀測誤差下，由雨滴譜儀模擬參數進行降雨公式(係數)之測試，結果顯示與統一公式(係數)相比，套用對應降雨類型公式(係數)後以R-(Z,ZDR)公式的改善最為明顯。而在雷達資料品管中使用自洽法修正Z偏差量與濕天線罩效應之衰減量，皆能明顯改善S與C波段雷達其降雨估計的表現。兩雷達於各個案降雨估計的比較，整體上 KDP參數之組合型公式(Hybrid)有最佳的表現；當兩雷達資料樣本數相近時，短波段雷達其KDP參數於降雨估計的優勢能明顯被凸顯，有較低的RRMSE與較高的相關係數。而後深入探討影響各降雨公式誤差表現之因子，包含ZDR觀測誤差造成降雨高估；濕天線罩效應致ZDR的誤差具空間分佈，使估計雨量圖有一明顯雨量不連續帶；以及Z參數公式受DSD隨高度變化影響對資料選取高度相對敏感等問題。在最後討論中顯示部分個案可透過合併兩雷達資料(即增加觀測資料數)而進一步提升降雨估計之準確度。
;To monitor extreme precipitation, the operational radar network in Taiwan will be set up with dual-polarimetric radars at S and C-band. It should be noted that each polarimetric parameter of different wavelength radars for quantitative precipitation estimation (QPE) has unique advantages and disadvantages. Thus, the goal of this study is to discuss and analyze QPE products of each radar as a base for merging rain maps.
Four different QPE relationships of each rain types, namely R-Z,R-KDP, R-(Z,ZDR) and R-(KDP,ZDR), are obtained from over-six-year NCU 2D-Video disdrometer data and applied to nine events observed by both RCWF S-band and NCU C-Pol radars from March of 2014 to August of 2015. The performances of radar-based QPE are investigated by comparing with 96 rain gauges.
Without considering radar measurement error, in the rain-type coefficients test by simulated radar parameters, it shows that most of algorithms improve after applied into the corresponding coefficients with respect to general coefficients, and R-(Z,ZDR) algorithm has the most significant improvement. In the radar data quality control process, the wet radome effect correction of Z has a positive impact on RCWF as well as NCU C-Pol radar. Overall,KDP-based relationships which combine with R-Z are the most accurate. In the comparable sampling frequency test, NCU C-Pol radar shows the advantage of KDP parameter for QPE at shorter wavelength with lower RRMSE and higher correlation coefficient. Besides, the factors resulting in QPE errors of each algorithm will be discussed, such as measurement error, wet radome effect and DSD variability with height. In the last part, the result indicates that when two radars’ data are included whenever available, the QPE performance can be further improved. |
