全球導航衛星系統無線電掩星（Global Navigation Satellite System Radio Occultation， GNSSRO）資料在觀測不足的區域提供豐富的資料，並在資料同化過程中扮演相當重要的角色。由於在低對流層的空氣密度劇烈變化，無線電波在掩星觀測裡會發生不同路徑的偏折，偏折角頻譜中的局部頻譜寬（Local Spectral Width， LSW）可用以代表無線電掩星觀測的不確定性，改善資料同化後的初始場進而修正熱帶氣旋的預報路徑。本研究採用格點統計插值（Gridpoint Statistical Interpolation， GSI）混成(hybrid)系統進行資料同化，並結合全球跨尺度天氣預報模式（Model for Prediction Across Scales–Atmosphere， MPAS），模擬2019年至2024年之間的熱帶氣旋預報。 GSI同化系統裡預設的觀測誤差與高度和緯度有著高度的相關性，利用LSW的特性部分取代原本預設的觀測誤差，同化過後再利用動態渦旋初始化（DVI）改善熱帶氣旋的初始場狀態，使其更接近最佳路徑的位置。 在資料同化循環過程中，在O-B和O-A的垂直剖面上BNDM1大部分比BND還要小，水平面上的分析增量相比於GTS實驗，BND和BNDM1都有很顯著的差異，而BND和BNDM1之間的差異在溫度與濕度增量上大多為顯著的正值。在預報路徑中，BND和BNDM1的路徑預報結果差異不大，但相對於GTS來說有些許的改善。 為了深入了解BND和BNDM1之間的差異，分別選取2019年米塔和2021年璨樹往後一天模擬結果表明，在路徑預報中，BNDM1比BND還更接近最佳路徑，在熱力場以及動力場上，LSW同化可以調整中低對流層的溫度、濕度以及風場，對於颱風的整體結構上，BNDM1可以很明確地顯示出靠近地形時颱風結構的變化。由系集敏感度分析(Ensemble Sensitivity Analysis， ESA)也表明BNDM1對於系集成員的風場調整也可以改善颱風預報的路徑。基於LSW品質控管的無線電掩星觀測資料為資料同化提供更好的初始條件，並提高熱帶氣旋路徑的預報度。;Global Positioning System (GPS) radio occultation (RO) data provides abundant observation data for assimilation system over regions with insufficient observation. The radio waves give rise to the different path deflection due to moisture variability in the lower troposphere. The local spectral width (LSW) from bending angle spectrum represents the uncertainty of RO observation， and then improves the initial fields for data assimilation to predict the track of the tropical cyclone. In this study， the Gridpoint Statistical Interpolation (GSI) hybrid system for data assimilation with the Model for Prediction Across Scales–Atmosphere (MPAS) is implemented and simulates the forecasts of tropical cyclone from 2019 to 2024. The representative error is highly correlated with altitude and latitude in the GSI system; however， the part of LSW is used to replace the default setting in GSI system for assimilating the RO data. The dynamical vortex initialization (DVI) is applied to improve the initial fields and get closer to the best track position after assimilation cycles. Four experiments are conducted: no data assimilation (CTRL)， assimilating conventional data only (GTS)， and RO bending angle with default (BND) or RO bending angle with LSW (BNDM1). During the data assimilation cycles， the analysis increments of BNDM1 on the vertical profile mostly get smaller values. The increments of BND and BNDM1 show the significant differences on the horizontal field compared to GTS. The differences between BND and BNDM1 in temperature and humidity increments are mostly significant positive values. The track of BNDM1 is similar to BND in most cases， but the tracks have statistically significant difference in all experiments. To understand the differences between BND and BNDM1， simulations were conducted one day later in Mitag (2019) and Chanthu (2021). The results showed the track of BNDM1 was closer to the best track than the track of BND. In thermal and dynamic structures， The data assimilation with LSW can effectively modulate the temperature， humidity， and wind field in the middle and lower troposphere. Regarding the overall structure of the typhoon， BNDM1 clearly demonstrates structural variations as the typhoon approaches terrain. Furthermore， the Ensemble Sensitivity Analysis (ESA) indicates that the adjustments made by BNDM1 to the wind fields of the ensemble members also contribute to the improvement of the simulated typhoon tracks. The results demonstrate RO data with LSW-based quality control provides better initial condition for data assimilation and improves the track prediction of tropical cyclone.
