ABSTRACT Global Position System (GPS) radio occultation (RO) soundings are assimilated to explore the impacts of GPS RO data on prediction of severe weather systems such as super cyclone Gonu (2007) over the northern Indian Ocean and Mei-yu frontal (2012) rainfall over Taiwan. A series of experiments were conducted using the advanced Weather Research and Forecasting (WRF) model with three dimensional variational method (3DVAR) to assimilate GPS RO refractivity from FORMOSAT-3/COSMIC (hereafter refer as GPS) and other datasets in order to verify the relative impact of GPS RO soundings. For the study of cyclone Gonu, we use Special Sensor Microwave Imager (SSM/I) retrieved precipitable water and near-surface oceanic wind speed, and radiosonde sounding (GTS), QuikSCAT derived winds, bogus vortex along with GPS RO data. Significant differences in cyclone track and intensity prediction are exhibited in various assimilations. GPS-cyc (GPS cyclic) outperformed well when compared to all other simulations till day 3 with the smallest cross-track error, SSM/I experiment obtains the strongest cyclone intensity. Sensitivity tests were also conducted to identify which GPS RO sounding plays a more crucial role in track prediction. It was found that the RO soundings in the vicinity of Gonu cyclone appear to modify the environmental conditions that result in a later development of a couplet of high and low pressure leading to an impact on track prediction, but the sounding closest to the cyclone center exhibits the largest contribution. Sensitivity experiments indicate that the retrieved GPS data information at upper levels with initial colder temperature increments indeed contributes iii more to improvement on track prediction. The improvement will not degrade until only the initial RO sounding information above 14 km height is retained in the model. Severe rainfall associated with a Mei-yu frontal system in Taiwan is studied using three different observational operators such as local bending angle, local refractivity, and non-local refractivity. The assimilations of the RO data are found beneficial to the forecast of severe rainfall associated with Mei-yu frontal system in Taiwan. Distributions of major rainfall in northern Taiwan on the second day, in particular, best benefit the performance of the local bending angle operator, followed by local refractivity operator. Both the operators appear to produce mostly larger positive moisture increments than the non-local operator. We found that the impact of RO data results from several critical RO observations near Mongolia and northern China. This study is the first attempt to provide an inter-comparison among the three RO operators, and show the feasibility of model assimilation with local bending angle.; 摘 要 全球定位系統掩星（GPS RO）觀測資料於本研究中被同化於模式中，進一 步想了解其對災害性天氣系統的影響，如北印度洋的超級氣旋Gonu（2007）和 通過台灣地區的梅雨鋒面（2012）。本研究一系列的實驗使用WRF(Weather Research and Forecasting )模式和三維變分資料同化方法（3DVAR），並於研究 中同化福爾摩沙衛星三號(FORMOSAT-3COSMIC)GPS RO折射率觀測資料和其 他用來驗證GPS RO的影響的觀測資料。 於颶風Gonu 的研究中， 我們使用SSM/I 衛星觀測資料(Special Sensor Microwave Imager)、傳統觀測資料（GTS）、QuikSCAT衛星觀測資料和包含GPS RO折射率觀測資料的虛擬渦漩。於本實驗中，同化不同的觀測資料在氣旋路徑 和強度預報有顯著不同，比較所有實驗路徑誤差，其中GPS-cyc(GPS cyclic)之氣 旋路徑誤差最小，而同化SSM/I衛星觀測資料模擬的氣旋強度最強。在敏感度實 驗中我們發現Gonu氣旋附近的GPS觀測資料修正了其環境條件，進一步造成對路 徑預測的影響，其中最接近氣旋中心的探空資料表現出最大的貢獻。敏感度實驗 中指出，GPS在模式高層初始場較冷的溫度增量對於路徑預測的改善有顯著的影 響，進一步確立同化GPS觀測資料於路徑預報中扮演著重要的角色。 梅雨（2012）鋒面系統數個實驗中，於模式中同化了GPS 掩星觀測資料並 使用三個不同的觀測算子如局地折射率(Local Reflectivity 簡稱REF)、局地偏折 角(Local Bending Angle 簡稱LBA)和非局地折射率(Nonlocal Refractivity 簡稱 EPH)，試圖了解同化不同的同化運算子對於梅雨預報的影響。我們發現同化掩 星資料對於梅雨(2012)個案預報結果皆有助益，其中LBA 實驗模擬第二天降雨 結果最接近觀測，其次為REF 實驗，這兩組實驗的模式初始場水氣增量皆比EPH 實驗大，其中我們發現數個位在蒙古和中國北方的掩星觀測資料扮演關鍵的角色。 在本研究中為國內首次嘗試這三種不同的掩星觀測算子間的比較，並表現模式同 化局地篇折角的可行性。
