| 摘要: | 本研究探討基於ZPHI與冪定律擬合(FIT)的兩種雙波長技術(Dual-wavelength technique)，在反演雲雨結構中液態水含量(Liquid water content, LWC)和雷達估計粒徑(Radar estimate size, RES)之能力。並建立一系列資料QC流程，測試其對反演結果不確定性之改善效益。這些雙波長技術利用從X波段和K波段雷達所得到的路徑總和衰減(Path-integrated attenuation, PIA)來估計每一格點之衰減率(Specific attenuation, A)。在理想模擬實驗中，本研究使用雨滴譜儀資料與背向散射模擬來獲得模擬觀測之雙波長回波場。實驗結果顯示，資料品質對反演結果影響甚鉅。評估FIT法反演之衰減率誤差，引入基礎ZPHI衰減修正與米氏散射(Mie scattering)修正，能使MAPE從32.3%下降至25.9%。比較不同反演方法之表現，無論LWC或RES場，ZPHI法皆較為貼合真實場的分布特徵，以LWC場為例，其擁有較高的相關係數(0.84)及較低的RMSE(0.11 g m-3)，而米氏散射修正流程能減少部分因電磁波背向散射強度減弱而導致的回波強度誤差。
在實際個案探討，本研究使用2021年宜蘭劇烈降雨觀測實驗(YESR 2021)與台灣區域豪雨觀測和預測實驗(TAHOPE)期間之TEAM-R與MRR-PRO雙波長觀測資料進行反演。並分別與雨滴譜儀模擬結果與NCAR S-Pol之雙偏極參數反演結果進行驗證。實驗結果顯示，得益於可在每個格點作調整的常數係數，ZPHI法反演之LWC-Z與RES-Z關係較為離散，在統計上與S-Pol雙偏參數反演結果更加吻合，其機率分布重疊比率分別為42及47%。使用本研究發展之資料同調及米氏散射修正等資料品管流程後，能有效減少反演結果之異常分布，其中又以FIT法結果之改善最為明顯。比較兩個案之降水特性差異，YESR反演之LWC量值較TAHOPE為高，RES則較小，反應冬季淺對流及梅雨期鋒面深對流之降水特性差異。
;This study investigates the capabilities of two dual-wavelength techniques, the ZPHI-based method and the power-law fitting (FIT) method, in retrieving liquid water content (LWC) and radar estimate size (RES) in cloud-rain regions. These techniques utilize path-integrated attenuation (PIA) obtained from X- and K-band radar to estimate specific attenuation (A) at each grid point. This study also establishes a series of data quality control (QC) processes to test the effectiveness in improving the uncertainty of retrievals. In the ideal simulation experiments, distrometer data and T-matrix backscattering simulations were used to generate simulated reflectivity fields. The results demonstrate the significant impact of data quality on the retrievals. Evaluating the A errors in the FIT method, the introduction of basic ZPHI attenuation correction and Mie scattering correction reduces the mean absolute percentage error (MAPE) from 32.3% to 25.9%. Comparing the performance of different retrieval methods, ZPHI method exhibit better agreement with the distribution characteristics of the truth for LWC and RES fields. For example, the LWC field shows a higher correlation coefficient of 0.84 and lower RMSE of 0.11 g m-3. The Mie scattering correction process reduces the error in reflectivity caused by the weakening of backscattering intensity.
In the case study, the retrieval was conducted using dual-wavelength observations from TEAM-R and MRR-PRO during the Yilan Extreme Storm Observation Experiment 2021 (YESR 2021) and Taiwan Heavy Rainfall Observation and Prediction Experiment (TAHOPE). The results were validated against simulated distrometer data and dual-pol retrievals from NCAR S-Pol RHI. The results indicate that, benefiting from adjustable constant coefficients at each grid point, the ZPHI method exhibits better statistical agreement with the S-Pol retrievals. The overlapping probability distribution ratios of the LWC-Z and RES-Z relationships is 42% and 47%, respectively. After applying the data synchronization and Mie scattering correction processes, the anomalous distribution in the retrievals is effectively reduced, with the most significant improvement observed in the FIT method. Comparing the precipitation characteristics between two cases, the LWC retrieved from YESR is higher than those from TAHOPE, while the RES is smaller. It reflects the differences in precipitation characteristics between shallow convective events during winter and deep convective events during the Mei-yu season. |
